apaam/mlpack
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge branch 'master' into cli-lsvm

Browse Source
This commit is contained in:
Yashwant Singh Parihar
2019-10-06 08:15:39 +05:30
committed by GitHub
parent 15042ace11 e172bcfdad
commit c906b6fbc2
251 changed files with 15452 additions and 2066 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.appveyor.yml
+19 -13
View File
@@ -11,9 +11,8 @@ environment:
boost_serialization-vc140.1.60.0.0/lib/native/address-model-64/lib/*.*"
BOOST_UNIT_TEST : "C:/projects/mlpack/\
boost_unit_test_framework-vc140.1.60.0.0/lib/native/address-model-64/lib/*.*"
ARMADILLO_DOWNLOAD : "http://sourceforge.net/projects/arma/files/\
armadillo-7.800.2.tar.xz"
ARMADILLO_LIBRARY : "C:/projects/mlpack/armadillo-7.800.2/\
ARMADILLO_DOWNLOAD : "http://ftp.fau.de/macports/distfiles/armadillo/armadillo-8.400.0.tar.xz"
ARMADILLO_LIBRARY : "C:/projects/mlpack/armadillo-8.400.0/\
build/Debug/armadillo.lib"
BLAS_LIBRARY : "%APPVEYOR_BUILD_FOLDER%/OpenBLAS.0.2.14.1/lib/native/lib/x64/\
libopenblas.dll.a"
@@ -25,8 +24,15 @@ environment:
matrix:
- APPVEYOR_BUILD_WORKER_IMAGE: Visual Studio 2015
VSVER: Visual Studio 14 2015 Win64
MSBUILD: C:\Program Files (x86)\MSBuild\14.0\bin\MSBuild.exe
- APPVEYOR_BUILD_WORKER_IMAGE: Visual Studio 2017
VSVER: Visual Studio 15 2017 Win64
MSBUILD: C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\MSBuild\15.0\Bin\MSBuild.exe
# Currently, the VS2019 build seems to always time out. This seems to be an
# AppVeyor issue.
# - APPVEYOR_BUILD_WORKER_IMAGE: Visual Studio 2019
# VSVER: Visual Studio 16 2019
# MSBUILD: C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\MSBuild\Current\Bin\MSBuild.exe
configuration: Release
@@ -62,28 +68,26 @@ build_script:
- ps: cp ${env:BOOST_UNIT_TEST} C:\projects\mlpack\boost_libs\
- echo TEST_ARMA is %ARMADILLO_DOWNLOAD%
- >
if not exist armadillo.tar.xz
appveyor DownloadFile %ARMADILLO_DOWNLOAD%
-FileName armadillo.tar.xz
- 7z x armadillo.tar.xz -so | 7z x -si -ttar > nul
- cd armadillo-7.800.2 && mkdir build && cd build
- 7z x armadillo.tar.xz -so -txz | 7z x -si -ttar > nul
- cd armadillo-8.400.0 && mkdir build && cd build
- >
cmake -G "Visual Studio 14 2015 Win64"
cmake -G "%VSVER%"
-DBLAS_LIBRARY:FILEPATH=%BLAS_LIBRARY%
-DLAPACK_LIBRARY:FILEPATH=%BLAS_LIBRARY%
-DCMAKE_PREFIX:FILEPATH="%APPVEYOR_BUILD_FOLDER%/armadillo"
-DBUILD_SHARED_LIBS=OFF
-DCMAKE_BUILD_TYPE=Release ..
- >
"C:\Program Files (x86)\MSBuild\14.0\Bin\MSBuild.exe"
"C:\projects\mlpack\armadillo-7.800.2\build\armadillo.sln"
"%MSBUILD%" "C:\projects\mlpack\armadillo-8.400.0\build\armadillo.sln"
/m /verbosity:quiet /p:Configuration=Release;Platform=x64
- cd C:\projects\mlpack && mkdir build && cd build
- >
cmake -G "Visual Studio 14 2015 Win64"
cmake -G "%VSVER%"
-DBLAS_LIBRARY:FILEPATH=%BLAS_LIBRARY%
-DLAPACK_LIBRARY:FILEPATH=%BLAS_LIBRARY%
-DARMADILLO_INCLUDE_DIR="C:/projects/mlpack/armadillo-7.800.2/include"
-DARMADILLO_INCLUDE_DIR="C:/projects/mlpack/armadillo-8.400.0/include"
-DARMADILLO_LIBRARY:FILEPATH=%ARMADILLO_LIBRARY%
-DBOOST_INCLUDEDIR:PATH=%BOOST_INCLUDE%
-DBOOST_LIBRARYDIR:PATH="C:/projects/mlpack/boost_libs"
@@ -92,8 +96,7 @@ build_script:
-DBUILD_PYTHON_BINDINGS=OFF
-DCMAKE_BUILD_TYPE=Release ..
- >
"C:\Program Files (x86)\MSBuild\14.0\Bin\MSBuild.exe"
"C:\projects\mlpack\build\mlpack.sln"
"%MSBUILD%" "C:\projects\mlpack\build\mlpack.sln"
/m /verbosity:minimal /nologo /p:BuildInParallel=true
/p:Configuration=Release;Platform=x64
@@ -143,6 +146,9 @@ build_script:
- ps: >
cp C:\projects\mlpack\doc\examples
C:\projects\mlpack\dist\win-installer\staging -recurse
- ps: >
cp C:\projects\mlpack\src\mlpack\tests\data\german.csv
C:\projects\mlpack\dist\win-installer\staging\examples\sample-ml-app\sample-ml-app\data\
# Checking current gitversion or mlpack version.
- ps: >
.gitignore
+1
View File
@@ -6,3 +6,4 @@ src/mlpack/core/util/arma_config.hpp
.idea
cmake-build-*
*.pyc
Testing/
.travis.yml
+7 -8
View File
@@ -1,4 +1,3 @@
sudo: required
language: cpp
matrix:
@@ -8,9 +7,9 @@ matrix:
env: CMAKE_OPTIONS="-DDEBUG=OFF -DPROFILE=OFF -DPYTHON_EXECUTABLE=/usr/bin/python"
before_install:
- sudo apt-get update
- sudo apt-get install -y --allow-unauthenticated libopenblas-dev liblapack-dev g++ libboost-all-dev python-pip cython python-numpy python-pandas
- sudo apt-get install -y --allow-unauthenticated libopenblas-dev liblapack-dev g++ libboost-all-dev python-pip cython python-numpy python-pandas xz-utils
- sudo pip install --upgrade --ignore-installed setuptools cython
- curl https://ftp.fau.de/macports/distfiles/armadillo/armadillo-6.500.5.tar.gz | tar xvz && cd armadillo*
- curl https://ftp.fau.de/macports/distfiles/armadillo/armadillo-8.400.0.tar.xz | tar -xvJ && cd armadillo*
- cmake . && make && sudo make install && cd ..
- sudo cp .travis/config.hpp /usr/include/armadillo_bits/config.hpp
@@ -19,9 +18,9 @@ matrix:
env: CMAKE_OPTIONS="-DDEBUG=OFF -DPROFILE=OFF -DPYTHON_EXECUTABLE=/usr/bin/python3"
before_install:
- sudo apt-get update
- sudo apt-get install -y --allow-unauthenticated libopenblas-dev liblapack-dev g++ libboost-all-dev python3-pip cython3 python3-numpy
- sudo apt-get install -y --allow-unauthenticated libopenblas-dev liblapack-dev g++ libboost-all-dev python3-pip cython3 python3-numpy xz-utils
- sudo pip3 install --upgrade --ignore-installed setuptools cython pandas
- curl https://ftp.fau.de/macports/distfiles/armadillo/armadillo-6.500.5.tar.gz | tar xvz && cd armadillo*
- curl https://ftp.fau.de/macports/distfiles/armadillo/armadillo-8.400.0.tar.xz | tar -xvJ && cd armadillo*
- cmake . && make && sudo make install && cd ..
- sudo cp .travis/config.hpp /usr/include/armadillo_bits/config.hpp
@@ -30,8 +29,8 @@ matrix:
env: CMAKE_OPTIONS="-DDEBUG=ON -DPROFILE=OFF -DBUILD_PYTHON_BINDINGS=OFF"
before_install:
- sudo apt-get update
- sudo apt-get install -y --allow-unauthenticated libopenblas-dev liblapack-dev g++ libboost-all-dev
- curl https://ftp.fau.de/macports/distfiles/armadillo/armadillo-6.500.5.tar.gz | tar xvz && cd armadillo*
- sudo apt-get install -y --allow-unauthenticated libopenblas-dev liblapack-dev g++ libboost-all-dev xz-utils
- curl https://ftp.fau.de/macports/distfiles/armadillo/armadillo-8.400.0.tar.xz | tar -xvJ && cd armadillo*
- cmake . && make && sudo make install && cd ..
- sudo cp .travis/config.hpp /usr/include/armadillo_bits/config.hpp
@@ -42,7 +41,7 @@ matrix:
- brew install openblas armadillo || brew install openblas armadillo
install:
- mkdir build && cd build && cmake $CMAKE_OPTIONS .. && travis_wait 60 make -j2
- mkdir build && cd build && cmake $CMAKE_OPTIONS .. && travis_wait 75 make -j2
script:
- CTEST_OUTPUT_ON_FAILURE=1 travis_wait 30 ctest -j2
.travis/config.hpp
+177 -49
View File
@@ -1,35 +1,65 @@
// Copyright (C) 2008-2012 NICTA (www.nicta.com.au)
// Copyright (C) 2008-2012 Conrad Sanderson
//
// This file is part of the Armadillo C++ library.
// It is provided without any warranty of fitness
// for any purpose. You can redistribute this file
// and/or modify it under the terms of the GNU
// Lesser General Public License (LGPL) as published
// by the Free Software Foundation, either version 3
// of the License or (at your option) any later version.
// (see http://www.opensource.org/licenses for more info)
// Copyright 2008-2016 Conrad Sanderson (http://conradsanderson.id.au)
// Copyright 2008-2016 National ICT Australia (NICTA)
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
// ------------------------------------------------------------------------
#if !defined(ARMA_USE_LAPACK)
#define ARMA_USE_LAPACK
//// Uncomment the above line if you have LAPACK or a high-speed replacement for LAPACK,
//// such as Intel's MKL, AMD's ACML, or the Accelerate framework.
//// Comment out the above line if you don't have LAPACK or a high-speed replacement for LAPACK,
//// such as Intel MKL, AMD ACML, or the Accelerate framework.
//// LAPACK is required for matrix decompositions (eg. SVD) and matrix inverse.
#endif
#if !defined(ARMA_USE_BLAS)
#define ARMA_USE_BLAS
//// Uncomment the above line if you have BLAS or a high-speed replacement for BLAS,
//// such as GotoBLAS, Intel's MKL, AMD's ACML, or the Accelerate framework.
//// Comment out the above line if you don't have BLAS or a high-speed replacement for BLAS,
//// such as OpenBLAS, GotoBLAS, Intel MKL, AMD ACML, or the Accelerate framework.
//// BLAS is used for matrix multiplication.
//// Without BLAS, matrix multiplication will still work, but might be slower.
#endif
#if !defined(ARMA_USE_NEWARP)
#define ARMA_USE_NEWARP
//// Uncomment the above line to enable the built-in partial emulation of ARPACK.
//// This is used for eigen decompositions of real (non-complex) sparse matrices, eg. eigs_sym(), svds()
#endif
#if !defined(ARMA_USE_ARPACK)
// #define ARMA_USE_ARPACK
//// Uncomment the above line if you have ARPACK or a high-speed replacement for ARPACK.
//// ARPACK is required for eigen decompositions of complex sparse matrices
#endif
#if !defined(ARMA_USE_SUPERLU)
// #define ARMA_USE_SUPERLU
//// Uncomment the above line if you have SuperLU.
//// SuperLU is used for solving sparse linear systems via spsolve()
//// Caveat: only SuperLU version 5.2 can be used!
#endif
#if !defined(ARMA_SUPERLU_INCLUDE_DIR)
// #define ARMA_SUPERLU_INCLUDE_DIR /usr/include/
//// If you're using SuperLU and want to explicitly include the SuperLU headers,
//// uncomment the above define and specify the appropriate include directory.
//// Make sure the directory has a trailing /
#endif
#define ARMA_USE_WRAPPER
//// Comment out the above line if you prefer to directly link with LAPACK and/or BLAS (eg. -llapack -lblas)
//// instead of linking indirectly with LAPACK and/or BLAS via Armadillo's run-time wrapper library.
//// Comment out the above line if you're getting linking errors when compiling your programs,
//// or if you prefer to directly link with LAPACK, BLAS + etc instead of the Armadillo runtime library.
//// You will then need to link your programs directly with -llapack -lblas instead of -larmadillo
// #define ARMA_BLAS_CAPITALS
//// Uncomment the above line if your BLAS and LAPACK libraries have capitalised function names (eg. ACML on 64-bit Windows)
@@ -45,31 +75,49 @@
//// Uncomment the above line if your BLAS and LAPACK libraries use "long long" instead of "int"
// #define ARMA_USE_TBB_ALLOC
//// Uncomment the above line if you want to use Intel TBB scalable_malloc() and scalable_free() instead of standard new[] and delete[]
//// Uncomment the above line if you want to use Intel TBB scalable_malloc() and scalable_free() instead of standard malloc() and free()
// #define ARMA_USE_MKL_ALLOC
//// Uncomment the above line if you want to use Intel MKL mkl_malloc() and mkl_free() instead of standard new[] and delete[]
//// Uncomment the above line if you want to use Intel MKL mkl_malloc() and mkl_free() instead of standard malloc() and free()
/* #undef ARMA_USE_ATLAS */
#define ARMA_ATLAS_INCLUDE_DIR /
// #define ARMA_USE_ATLAS
// #define ARMA_ATLAS_INCLUDE_DIR /usr/include/
//// If you're using ATLAS and the compiler can't find cblas.h and/or clapack.h
//// uncomment the above define and specify the appropriate include directory.
//// Make sure the directory has a trailing /
#define ARMA_64BIT_WORD
//// Uncomment the above line if you require matrices/vectors capable of holding more than 4 billion elements.
//// Your machine and compiler must have support for 64 bit integers (eg. via "long" or "long long")
#if !defined(ARMA_USE_CXX11)
#define ARMA_USE_CXX11
//// Uncomment the above line if you have a C++ compiler that supports the C++11 standard
//// This will enable additional features, such as use of initialiser lists
//// Uncomment the above line to forcefully enable use of C++11 features (eg. initialiser lists).
//// Note that ARMA_USE_CXX11 is automatically enabled when a C++11 compiler is detected.
#endif
#if !defined(ARMA_USE_OPENMP)
// #define ARMA_USE_OPENMP
//// Uncomment the above line to forcefully enable use of OpenMP for parallelisation.
//// Note that ARMA_USE_OPENMP is automatically enabled when a compiler supporting OpenMP 3.1 is detected.
#endif
#if !defined(ARMA_64BIT_WORD)
#define ARMA_64BIT_WORD
//// Uncomment the above line if you require matrices/vectors capable of holding more than 4 billion elements.
//// Your machine and compiler must have support for 64 bit integers (eg. via "long" or "long long").
//// Note that ARMA_64BIT_WORD is automatically enabled when a C++11 compiler is detected.
#endif
#if !defined(ARMA_USE_HDF5)
/* #undef ARMA_USE_HDF5 */
//// Uncomment the above line if you want the ability to save and load matrices stored in the HDF5 format;
//// the hdf5.h header file must be available on your system and you will need to link with the hdf5 library (eg. -lhdf5)
// #define ARMA_USE_HDF5
//// Uncomment the above line to allow the ability to save and load matrices stored in HDF5 format;
//// the hdf5.h header file must be available on your system,
//// and you will need to link with the hdf5 library (eg. -lhdf5)
#endif
// #define ARMA_USE_HDF5_ALT
#if defined(ARMA_USE_HDF5_ALT) && defined(ARMA_USE_WRAPPER)
#undef ARMA_USE_HDF5
#define ARMA_USE_HDF5
// #define ARMA_HDF5_INCLUDE_DIR /usr/include/
#endif
#if !defined(ARMA_MAT_PREALLOC)
@@ -80,6 +128,18 @@
//// If you mainly use lots of very small vectors (eg. <= 4 elements),
//// change the number to the size of your vectors.
#if !defined(ARMA_OPENMP_THRESHOLD)
#define ARMA_OPENMP_THRESHOLD 320
#endif
//// The minimum number of elements in a matrix to allow OpenMP based parallelisation;
//// it must be an integer that is at least 1.
#if !defined(ARMA_OPENMP_THREADS)
#define ARMA_OPENMP_THREADS 10
#endif
//// The maximum number of threads to use for OpenMP based parallelisation;
//// it must be an integer that is at least 1.
#if !defined(ARMA_SPMAT_CHUNKSIZE)
#define ARMA_SPMAT_CHUNKSIZE 256
#endif
@@ -99,27 +159,39 @@
//// This is mainly useful for debugging of the library.
// #define ARMA_USE_BOOST
// #define ARMA_USE_BOOST_DATE
#if !defined(ARMA_DEFAULT_OSTREAM)
#define ARMA_DEFAULT_OSTREAM std::cout
#if defined(ARMA_DEFAULT_OSTREAM)
#pragma message ("WARNING: support for ARMA_DEFAULT_OSTREAM is deprecated and will be removed;")
#pragma message ("WARNING: use ARMA_COUT_STREAM and ARMA_CERR_STREAM instead")
#endif
#define ARMA_PRINT_LOGIC_ERRORS
#define ARMA_PRINT_RUNTIME_ERRORS
//#define ARMA_PRINT_HDF5_ERRORS
#define ARMA_HAVE_STD_ISFINITE
#define ARMA_HAVE_STD_ISINF
#define ARMA_HAVE_STD_ISNAN
#define ARMA_HAVE_STD_SNPRINTF
#if !defined(ARMA_COUT_STREAM)
#if defined(ARMA_DEFAULT_OSTREAM)
// for compatibility with earlier versions of Armadillo
#define ARMA_COUT_STREAM ARMA_DEFAULT_OSTREAM
#else
#define ARMA_COUT_STREAM std::cout
#endif
#endif
#define ARMA_HAVE_LOG1P
#define ARMA_HAVE_GETTIMEOFDAY
#if !defined(ARMA_CERR_STREAM)
#if defined(ARMA_DEFAULT_OSTREAM)
// for compatibility with earlier versions of Armadillo
#define ARMA_CERR_STREAM ARMA_DEFAULT_OSTREAM
#else
#define ARMA_CERR_STREAM std::cerr
#endif
#endif
#if !defined(ARMA_PRINT_ERRORS)
#define ARMA_PRINT_ERRORS
//// Comment out the above line if you don't want errors and warnings printed (eg. failed decompositions)
#endif
#if !defined(ARMA_PRINT_HDF5_ERRORS)
// #define ARMA_PRINT_HDF5_ERRORS
#endif
#if defined(ARMA_DONT_USE_LAPACK)
#undef ARMA_USE_LAPACK
@@ -129,15 +201,71 @@
#undef ARMA_USE_BLAS
#endif
#if defined(ARMA_DONT_USE_NEWARP) || !defined(ARMA_USE_LAPACK)
#undef ARMA_USE_NEWARP
#endif
#if defined(ARMA_DONT_USE_ARPACK)
#undef ARMA_USE_ARPACK
#endif
#if defined(ARMA_DONT_USE_SUPERLU)
#undef ARMA_USE_SUPERLU
#undef ARMA_SUPERLU_INCLUDE_DIR
#endif
#if defined(ARMA_DONT_USE_ATLAS)
#undef ARMA_USE_ATLAS
#undef ARMA_ATLAS_INCLUDE_DIR
#endif
#if defined(ARMA_DONT_PRINT_LOGIC_ERRORS)
#undef ARMA_PRINT_LOGIC_ERRORS
#if defined(ARMA_DONT_USE_WRAPPER)
#undef ARMA_USE_WRAPPER
#undef ARMA_USE_HDF5_ALT
#endif
#if defined(ARMA_DONT_PRINT_RUNTIME_ERRORS)
#undef ARMA_PRINT_RUNTIME_ERRORS
#if defined(ARMA_DONT_USE_CXX11)
#undef ARMA_USE_CXX11
#undef ARMA_USE_EXTERN_CXX11_RNG
#endif
#if defined(ARMA_DONT_USE_OPENMP)
#undef ARMA_USE_OPENMP
#endif
#if defined(ARMA_USE_WRAPPER)
#if defined(ARMA_USE_CXX11)
#if !defined(ARMA_USE_EXTERN_CXX11_RNG)
// #define ARMA_USE_EXTERN_CXX11_RNG
#endif
#endif
#endif
#if defined(ARMA_DONT_USE_EXTERN_CXX11_RNG)
#undef ARMA_USE_EXTERN_CXX11_RNG
#endif
#if defined(ARMA_32BIT_WORD)
#undef ARMA_64BIT_WORD
#endif
#if defined(ARMA_DONT_USE_HDF5)
#undef ARMA_USE_HDF5
#undef ARMA_USE_HDF5_ALT
#endif
#if defined(ARMA_DONT_PRINT_ERRORS)
#undef ARMA_PRINT_ERRORS
#endif
#if defined(ARMA_DONT_PRINT_HDF5_ERRORS)
#undef ARMA_PRINT_HDF5_ERRORS
#endif
// if Armadillo was installed on this system via CMake and ARMA_USE_WRAPPER is not defined,
// ARMA_AUX_LIBS lists the libraries required by Armadillo on this system, and
// ARMA_AUX_INCDIRS lists the include directories required by Armadillo on this system.
// Do not use these unless you know what you are doing.
#define ARMA_AUX_LIBS
#define ARMA_AUX_INCDIRS
CMake/FindEnsmallen.cmake
+1 -1
View File
@@ -52,7 +52,7 @@ endif ()
# Checks 'REQUIRED', 'QUIET' and versions.
include(FindPackageHandleStandardArgs)
find_package_handle_standard_args(ensmallen
find_package_handle_standard_args(Ensmallen
REQUIRED_VARS ENSMALLEN_INCLUDE_DIR
VERSION_VAR ENSMALLEN_VERSION_STRING)
CMake/FindStbImage.cmake
+23
View File
@@ -0,0 +1,23 @@
# - Find STB_IMAGE
# Find the STB_IMAGE C++ library
#
# This module sets the following variables:
# STB_IMAGE_FOUND - set to true if the library is found
# STB_IMAGE_INCLUDE_DIR - list of required include directories
file(GLOB STB_IMAGE_SEARCH_PATHS
${CMAKE_BINARY_DIR}/deps/stb)
find_path(STB_IMAGE_INCLUDE_DIR
NAMES stb_image.h stb_image_write.h
PATHS ${STB_IMAGE_SEARCH_PATHS})
if(STB_IMAGE_INCLUDE_DIR)
set(STB_IMAGE_FOUND YES)
endif ()
# Checks 'REQUIRED'.
include(FindPackageHandleStandardArgs)
find_package_handle_standard_args(STB_IMAGE
REQUIRED_VARS STB_IMAGE_INCLUDE_DIR)
mark_as_advanced(STB_IMAGE_INCLUDE_DIR)
CMake/RunProgram.cmake
+6 -1
View File
@@ -5,4 +5,9 @@
#
# PROGRAM: the program to run to.
# OUTPUT_FILE: the file to store the output in.
execute_process(COMMAND ${PROGRAM} OUTPUT_FILE ${OUTPUT_FILE})
execute_process(COMMAND ${PROGRAM} OUTPUT_FILE ${OUTPUT_FILE}
ERROR_VARIABLE err)
if (err)
message(FATAL_ERROR "Fatal error running ${PROGRAM}: ${err}!")
endif ()
CMakeLists.txt
+56 -3
View File
@@ -13,6 +13,7 @@ option(TEST_VERBOSE "Run test cases with verbose output." OFF)
option(BUILD_TESTS "Build tests." ON)
option(BUILD_CLI_EXECUTABLES "Build command-line executables." ON)
option(DOWNLOAD_ENSMALLEN "If ensmallen is not found, download it." ON)
option(DOWNLOAD_STB_IMAGE "Download stb_image for image loading." ON)
# Currently Python bindings aren't known to build successfully on Windows, so
# set BUILD_PYTHON_BINDINGS to OFF when the platform is Windows.
@@ -75,6 +76,10 @@ if(CMAKE_COMPILER_IS_GNUCC OR "${CMAKE_CXX_COMPILER_ID}" STREQUAL "Clang")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -Wextra -ftemplate-depth=1000")
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wall -Wextra")
# To remove unused functions warnings.
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-unused-function")
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wno-unused-function")
endif()
# These support libraries are used if we need to link against something
@@ -233,7 +238,7 @@ endif()
# ENSMALLEN_INCLUDE_DIR - include directory for ensmallen
# MATHJAX_ROOT - root of MathJax installation
find_package(Armadillo 6.500.0 REQUIRED)
find_package(Armadillo 8.400.0 REQUIRED)
# If Armadillo was compiled without ARMA_64BIT_WORD and we are on a 64-bit
# system (where size_t will be 64 bits), suggest to the user that they should
@@ -310,15 +315,62 @@ endif ()
set(MLPACK_INCLUDE_DIRS ${MLPACK_INCLUDE_DIRS} ${ARMADILLO_INCLUDE_DIRS})
set(MLPACK_LIBRARIES ${MLPACK_LIBRARIES} ${ARMADILLO_LIBRARIES})
# Find stb_image.h and stb_image_write.h.
find_package(StbImage)
# Download stb_image for image loading.
if (NOT STB_IMAGE_FOUND)
if (DOWNLOAD_STB_IMAGE)
set(STB_DIR "stb")
install(DIRECTORY DESTINATION ${CMAKE_BINARY_DIR}/deps/${STB_DIR})
file(DOWNLOAD http://mlpack.org/files/stb/stb_image.h
"${CMAKE_BINARY_DIR}/deps/${STB_DIR}/stb_image.h"
STATUS STB_IMAGE_DOWNLOAD_STATUS_LIST LOG STB_IMAGE_DOWNLOAD_LOG
SHOW_PROGRESS)
list(GET STB_IMAGE_DOWNLOAD_STATUS_LIST 0 STB_IMAGE_DOWNLOAD_STATUS)
file(DOWNLOAD http://mlpack.org/files/stb/stb_image_write.h
"${CMAKE_BINARY_DIR}/deps/${STB_DIR}/stb_image_write.h"
STATUS STB_IMAGE_WRITE_DOWNLOAD_STATUS_LIST
LOG STB_IMAGE_WRITE_DOWNLOAD_LOG
SHOW_PROGRESS)
list(GET STB_IMAGE_WRITE_DOWNLOAD_STATUS_LIST 0
STB_IMAGE_WRITE_DOWNLOAD_STATUS)
if (STB_IMAGE_DOWNLOAD_STATUS EQUAL 0 AND
STB_IMAGE_WRITE_DOWNLOAD_STATUS EQUAL 0)
set(MLPACK_INCLUDE_DIRS ${MLPACK_INCLUDE_DIRS}
"${CMAKE_BINARY_DIR}/deps/${STB_DIR}/")
message(STATUS
"Successfully downloaded stb into ${CMAKE_BINARY_DIR}/deps/${STB_DIR}/")
# Now we have to also ensure these header files get installed.
install(FILES ${CMAKE_BINARY_DIR}/deps/${STB_DIR}/stb_image.h DESTINATION ${CMAKE_INSTALL_INCLUDEDIR})
install(FILES ${CMAKE_BINARY_DIR}/deps/${STB_DIR}/stb_image_write.h DESTINATION ${CMAKE_INSTALL_INCLUDEDIR})
add_definitions(-DHAS_STB)
else ()
list(GET STB_IMAGE_DOWNLOAD_STATUS_LIST 1 STB_DOWNLOAD_ERROR)
message(WARNING
"Could not download stb! Error code ${STB_DOWNLOAD_STATUS}: ${STB_DOWNLOAD_ERROR}! Error log: ${STB_DOWNLOAD_LOG}")
message(WARNING
"stb/stb_image.h is not installed. Image utilities will not be available!")
endif ()
else ()
message(WARNING
"stb/stb_image.h is not installed. Image utilities will not be available!")
endif ()
else ()
# Already has STB installed.
add_definitions(-DHAS_STB)
set(MLPACK_INCLUDE_DIRS ${MLPACK_INCLUDE_DIRS} ${STB_IMAGE_INCLUDE_DIR})
endif ()
# Find ensmallen.
# Once ensmallen is readily available in package repos, the automatic downloader
# here can be removed.
find_package(Ensmallen 1.10.0)
find_package(Ensmallen 2.10.0)
if (NOT ENSMALLEN_FOUND)
if (DOWNLOAD_ENSMALLEN)
file(DOWNLOAD http://www.ensmallen.org/files/ensmallen-latest.tar.gz
"${CMAKE_BINARY_DIR}/deps/ensmallen-latest.tar.gz"
STATUS ENS_DOWNLOAD_STATUS_LIST LOG ENS_DOWBLOAD_LOG
STATUS ENS_DOWNLOAD_STATUS_LIST LOG ENS_DOWNLOAD_LOG
SHOW_PROGRESS)
list(GET ENS_DOWNLOAD_STATUS_LIST 0 ENS_DOWNLOAD_STATUS)
if (ENS_DOWNLOAD_STATUS EQUAL 0)
@@ -329,6 +381,7 @@ if (NOT ENSMALLEN_FOUND)
# Get the name of the directory.
file (GLOB ENS_DIRECTORIES RELATIVE "${CMAKE_BINARY_DIR}/deps/"
"${CMAKE_BINARY_DIR}/deps/ensmallen-[0-9]*.[0-9]*.[0-9]*")
list(FILTER ENS_DIRECTORIES EXCLUDE REGEX "ensmallen-.*\.tar\.gz")
list(LENGTH ENS_DIRECTORIES ENS_DIRECTORIES_LEN)
if (ENS_DIRECTORIES_LEN EQUAL 1)
list(GET ENS_DIRECTORIES 0 ENSMALLEN_INCLUDE_DIR)
CODE_OF_CONDUCT.md
+96
View File
@@ -0,0 +1,96 @@
# mlpack Code of Conduct
In the interest of fostering an open and welcoming environment, we as
contributors and maintainers pledge to making participation in our project and
our community a harassment-free experience for everyone, regardless of age, body
size, disability, ethnicity, sex characteristics, gender identity and
expression, level of experience, education, socio-economic status, nationality,
personal appearance, race, religion, or sexual identity and orientation.
## Our Standards
Examples of behavior that contributes to creating a positive environment
include:
* Using welcoming and inclusive language
* Being respectful of differing viewpoints and experiences
* Gracefully accepting constructive criticism
* Showing empathy towards other community members
Examples of unacceptable behavior by participants include:
* The use of sexualized language or imagery and unwelcome sexual attention or
advances
* Trolling, insulting/derogatory comments, and personal or political attacks
* Public or private harassment
* Publishing others' private information, such as a physical or electronic
address, without explicit permission
## Our Responsibilities
Project maintainers are responsible for clarifying the standards of acceptable
behavior and are expected to take appropriate and fair corrective action in
response to any instances of unacceptable behavior.
Project maintainers have the right and responsibility to remove, edit, or
reject comments, commits, code, wiki edits, issues, and other contributions
that are not aligned to this Code of Conduct, or to ban temporarily or
permanently any contributor for other behaviors that they deem inappropriate,
threatening, offensive, or harmful.
## Scope
This Code of Conduct applies within all project spaces, and it also applies when
an individual is representing the project or its community in public spaces.
Examples of representing a project or community include using an official
project e-mail address, posting via an official social media account, or acting
as an appointed representative at an online or offline event. Representation of
a project may be further defined and clarified by project maintainers.
## Enforcement
Instances of abusive, harassing, or otherwise unacceptable behavior may be
reported by contacting the project team at conduct@mlpack.org. All
complaints will be reviewed and investigated and will result in a response that
is deemed necessary and appropriate to the circumstances. The project team is
obligated to maintain confidentiality with regard to the reporter of an incident.
Further details of specific enforcement policies may be posted separately.
Project maintainers who do not follow or enforce the Code of Conduct in good
faith may face temporary or permanent repercussions as determined by other
members of the project's leadership.
## Reporting
If you believe someone is violating the code of conduct we ask that you report
it by emailing conduct@mlpack.org. All reports will be kept confidential. In
some cases we may determine that a public statement will need to be made. If
that's the case, the identities of all victims and reporters will remain
confidential unless those individuals instruct us otherwise.
If you are unsure whether the incident is a violation, or whether the space
where it happened is covered by this Code of Conduct, we encourage you to still
report it. We would much rather have a few extra reports where we decide to take
no action, rather than miss a report of an actual violation. We do not look
negatively on you if we find the incident is not a violation. And knowing about
incidents that are not violations, or happen outside our spaces, can also help
us to improve the Code of Conduct or the processes surrounding it.
In your report please include:
* Your contact info (so we can get in touch with you if we need to follow up)
* Names (real, nicknames, or pseudonyms) of any individuals involved. If there
were other witnesses besides you, please try to include them as well.
* When and where the incident occurred. Please be as specific as possible.
* Your account of what occurred. If there is a publicly available record
(e.g. a mailing list archive or a public IRC logger) please include a link.
* Any extra context you believe existed for the incident.
* If you believe this incident is ongoing.
* Any other information you believe we should have.
## Attribution
This Code of Conduct is adapted from the [Contributor Covenant][homepage],
version 1.4, available at
https://www.contributor-covenant.org/version/1/4/code-of-conduct.html, and
includes some aspects of the Drupal Code of Conduct.
COPYRIGHT.txt
+4
View File
@@ -113,6 +113,10 @@ Copyright:
Copyright 2019, Dan Timson
Copyright 2019, Miguel Canteras <mcanteras@gmail.com>
Copyright 2019, Bishwa Karki <karkeebishwa1@gmail.com>
Copyright 2019, Mehul Kumar Nirala <mehulkumarnirala@gmail.com>
Copyright 2019, Yashwant Singh Parihar <yashwantsingh.sngh@gmail.com>
Copyright 2019, Heet Sankesara <heetsankesara3@gmail.com>
Copyright 2019, Jeffin Sam <sam.jeffin@gmail.com>
License: BSD-3-clause
All rights reserved.
Doxyfile
+1 -1
View File
@@ -4,7 +4,7 @@
# Project related configuration options
#---------------------------------------------------------------------------
PROJECT_NAME = mlpack
PROJECT_NUMBER = 3.1.0
PROJECT_NUMBER = 3.1.1
OUTPUT_DIRECTORY = ./doc
CREATE_SUBDIRS = NO
OUTPUT_LANGUAGE = English
HISTORY.md
+1 -2
View File
@@ -666,5 +666,4 @@
* Initial release. See any resolved tickets numbered less than #196 or
execute this query:
http://www.mlpack.org/trac/query?status=closed&milestone=mlpack+1.0.0
LICENSE.txt
+4
View File
@@ -12,6 +12,10 @@ licensed under the Boost Software License, version 1.0. This code is found in
src/mlpack/core/boost_backport/ and more details on the licensing are available
there.
mlpack may contain some usage of the source code of stb, which is licensed
under the MIT License and the Public Domain (www.unlicense.org). This code
is used in src/mlpack/core/data/load_image.hpp.
----
Copyright (c) 2007-2018, mlpack contributors (see COPYRIGHT.txt)
All rights reserved.
README.md
+8 -2
View File
@@ -23,7 +23,7 @@ src="https://cdn.rawgit.com/mlpack/mlpack.org/e7d36ed8/mlpack-black.svg" style="
<p align="center">
<em>
Download:
<a href="https://www.mlpack.org/files/mlpack-3.1.1.tar.gz">current stable version (3.1.1)</a>
<a href="https://www.mlpack.org/files/mlpack-3.2.1.tar.gz">current stable version (3 2.1)</a>
</em>
</p>
@@ -83,10 +83,11 @@ Citations are beneficial for the growth and improvement of mlpack.
mlpack has the following dependencies:
Armadillo >= 6.500.0
Armadillo >= 8.400.0
Boost (program_options, math_c99, unit_test_framework, serialization,
spirit)
CMake >= 3.3.2
ensmallen >= 2.10.0
All of those should be available in your distribution's package manager. If
not, you will have to compile each of them by hand. See the documentation for
@@ -100,6 +101,9 @@ following Python packages are installed:
numpy
pandas >= 0.15.0
If the STB library headers are available, image loading support will be
compiled.
If you are compiling Armadillo by hand, ensure that LAPACK and BLAS are enabled.
### 4. Building mlpack from source
@@ -160,10 +164,12 @@ Options are specified with the -D flag. The allowed options include:
ARMADILLO_LIBRARY=(/path/to/armadillo/libarmadillo.so): Armadillo library
BUILD_CLI_EXECUTABLES=(ON/OFF): whether or not to build command-line programs
BUILD_PYTHON_BINDINGS=(ON/OFF): whether or not to build Python bindings
PYTHON_EXECUTABLE=(/path/to/python_version): Path to specific Python executable
BUILD_TESTS=(ON/OFF): whether or not to build tests
BUILD_SHARED_LIBS=(ON/OFF): compile shared libraries as opposed to
static libraries
DOWNLOAD_ENSMALLEN=(ON/OFF): If ensmallen is not found, download it
DOWNLOAD_STB_IMAGE=(ON/OFF): If STB is not found, download it
ENSMALLEN_INCLUDE_DIR=(/path/to/ensmallen/include): path to include directory
for ensmallen
USE_OPENMP=(ON/OFF): whether or not to use OpenMP if available
doc/examples/sample-ml-app/sample-ml-app/sample-ml-app.vcxproj
+4 -4
View File
@@ -104,16 +104,16 @@
<SDLCheck>true</SDLCheck>
<PreprocessorDefinitions>_DEBUG;_CONSOLE;%(PreprocessorDefinitions)</PreprocessorDefinitions>
<ConformanceMode>false</ConformanceMode>
<AdditionalIncludeDirectories>C:\boost\boost_1_66_0;C:\mlpack\armadillo-8.500.1\include;C:\mlpack\mlpack-3.1.1\build\include;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
<AdditionalIncludeDirectories>C:\boost\boost_1_66_0;C:\mlpack\armadillo-8.500.1\include;C:\mlpack\mlpack-3.2.1\build\include;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
</ClCompile>
<Link>
<SubSystem>Console</SubSystem>
<GenerateDebugInformation>true</GenerateDebugInformation>
<AdditionalDependencies>C:\mlpack\mlpack-3.1.1\build\Debug\mlpack.lib;C:\boost\boost_1_66_0\lib64-msvc-14.1\libboost_serialization-vc141-mt-gd-x64-1_66.lib;C:\boost\boost_1_66_0\lib64-msvc-14.1\libboost_program_options-vc141-mt-gd-x64-1_66.lib;%(AdditionalDependencies)</AdditionalDependencies>
<AdditionalDependencies>C:\mlpack\mlpack-3.2.1\build\Debug\mlpack.lib;C:\boost\boost_1_66_0\lib64-msvc-14.1\libboost_serialization-vc141-mt-gd-x64-1_66.lib;C:\boost\boost_1_66_0\lib64-msvc-14.1\libboost_program_options-vc141-mt-gd-x64-1_66.lib;%(AdditionalDependencies)</AdditionalDependencies>
</Link>
<PostBuildEvent>
<Command>xcopy /y "C:\mlpack\mlpack-3.1.1\build\Debug\mlpack.dll" $(OutDir)
xcopy /y "C:\mlpack\mlpack-3.1.1\packages\OpenBLAS.0.2.14.1\lib\native\bin\x64\*.dll" $(OutDir)
<Command>xcopy /y "C:\mlpack\mlpack-3.2.1\build\Debug\mlpack.dll" $(OutDir)
xcopy /y "C:\mlpack\mlpack-3.2.1\packages\OpenBLAS.0.2.14.1\lib\native\bin\x64\*.dll" $(OutDir)
xcopy /y "$(ProjectDir)..\..\..\..\src\mlpack\tests\data\german.csv" "$(ProjectDir)data\german.csv*"</Command>
</PostBuildEvent>
</ItemDefinitionGroup>
doc/guide/build.hpp
+20 -10
View File
@@ -29,7 +29,7 @@ to build mlpack on Windows, see \ref build_windows (alternatively, you can read
is based on older versions).
You can download the latest mlpack release from here:
<a href="https://www.mlpack.org/files/mlpack-3.1.1.tar.gz">mlpack-3.1.1</a>
<a href="https://www.mlpack.org/files/mlpack-3.2.1.tar.gz">mlpack-3.2.1</a>
@section build_simple Simple Linux build instructions
@@ -37,9 +37,9 @@ Assuming all dependencies are installed in the system, you can run the commands
below directly to build and install mlpack.
@code
$ wget https://www.mlpack.org/files/mlpack-3.1.1.tar.gz
$ tar -xvzpf mlpack-3.1.1.tar.gz
$ mkdir mlpack-3.1.1/build && cd mlpack-3.1.1/build
$ wget https://www.mlpack.org/files/mlpack-3.2.1.tar.gz
$ tar -xvzpf mlpack-3.2.1.tar.gz
$ mkdir mlpack-3.2.1/build && cd mlpack-3.2.1/build
$ cmake ../
$ make -j4 # The -j is the number of cores you want to use for a build.
$ sudo make install
@@ -64,8 +64,8 @@ configure mlpack.
First we should unpack the mlpack source and create a build directory.
@code
$ tar -xvzpf mlpack-3.1.1.tar.gz
$ cd mlpack-3.1.1
$ tar -xvzpf mlpack-3.2.1.tar.gz
$ cd mlpack-3.2.1
$ mkdir build
@endcode
@@ -76,9 +76,15 @@ The directory can have any name, not just 'build', but 'build' is sufficient.
mlpack depends on the following libraries, which need to be installed on the
system and have headers present:
- Armadillo >= 6.500.0 (with LAPACK support)
- Armadillo >= 8.400.0 (with LAPACK support)
- Boost (math_c99, program_options, serialization, unit_test_framework, heap,
spirit) >= 1.49
- ensmallen >= 2.10.0 (will be downloaded if not found)
In addition, mlpack has the following optional dependencies:
- STB: this will allow loading of images; the library is downloaded if not
found and the CMake variable DOWNLOAD_STB_IMAGE is set to ON (the default)
For Python bindings, the following packages are required:
@@ -93,7 +99,8 @@ In Ubuntu and Debian, you can get all of these dependencies through apt:
@code
# apt-get install libboost-math-dev libboost-program-options-dev
libboost-test-dev libboost-serialization-dev libarmadillo-dev binutils-dev
python-pandas python-numpy cython python-setuptools
python-pandas python-numpy cython python-setuptools libensmallen-dev
libstb-dev
@endcode
On Fedora, Red Hat, or CentOS, these same dependencies can be obtained via dnf:
@@ -101,11 +108,12 @@ On Fedora, Red Hat, or CentOS, these same dependencies can be obtained via dnf:
@code
# dnf install boost-devel boost-test boost-program-options boost-math
armadillo-devel binutils-devel python2-Cython python2-setuptools
python2-numpy python2-pandas
python2-numpy python2-pandas ensmallen-devel stbi-devel
@endcode
(It's also possible to use python3 packages from the package manager---mlpack
will work with either.)
will work with either. Also, the ensmallen-devel package is only available in
Fedora 29 or RHEL7 or newer.)
@section build_config Configuring CMake
@@ -146,8 +154,10 @@ The full list of options mlpack allows:
(default OFF)
- DOWNLOAD_ENSMALLEN=(ON/OFF): If ensmallen is not found, download it
(default ON)
- DOWNLOAD_STB_IMAGE=(ON/OFF): If STB is not found, download it (default ON)
- BUILD_WITH_COVERAGE=(ON/OFF): Build with support for code coverage tools
(gcc only) (default OFF)
- PYTHON_EXECUTABLE=(/path/to/python_version): Path to specific Python executable
- BUILD_MARKDOWN_BINDINGS=(ON/OFF): Build Markdown bindings for website
documentation (default OFF)
- MATHJAX=(ON/OFF): use MathJax for generated Doxygen documentation (default
doc/guide/python_quickstart.hpp
+5 -5
View File
@@ -17,11 +17,11 @@ Installing the mlpack bindings for Python is straightforward. It's easy to use
conda or pip to do this:
@code{.sh}
pip install mlpack/mlpack3
pip install mlpack
@endcode
@code{.sh}
conda install -c mlpack mlpack
conda install -c conda-forge mlpack
@endcode
Otherwise, we can build the Python bindings from scratch, as follows. First we
@@ -31,9 +31,9 @@ build and install mlpack. You can copy-paste the commands into your shell.
@code{.sh}
sudo apt-get install libboost-all-dev g++ cmake libarmadillo-dev python-pip wget
sudo pip install cython setuptools distutils numpy pandas
wget https://www.mlpack.org/files/mlpack-3.1.1.tar.gz
tar -xvzpf mlpack-3.1.1.tar.gz
mkdir -p mlpack-3.1.1/build/ && cd mlpack-3.1.1/build/
wget https://www.mlpack.org/files/mlpack-3.2.1.tar.gz
tar -xvzpf mlpack-3.2.1.tar.gz
mkdir -p mlpack-3.2.1/build/ && cd mlpack-3.2.1/build/
cmake ../ && make -j4 && sudo make install
@endcode
doc/guide/sample_ml_app.hpp
+4 -4
View File
@@ -29,18 +29,18 @@ mlpack and dependencies in Release Mode).
@code
- C:\boost\boost_1_66_0
- C:\mlpack\armadillo-8.500.1\include
- C:\mlpack\mlpack-3.1.1\build\include
- C:\mlpack\mlpack-3.2.1\build\include
@endcode
- Under Linker > Input > Additional Dependencies add:
@code
- C:\mlpack\mlpack-3.1.1\build\Debug\mlpack.lib
- C:\mlpack\mlpack-3.2.1\build\Debug\mlpack.lib
- C:\boost\boost_1_66_0\lib64-msvc-14.1\libboost_serialization-vc141-mt-gd-x64-1_66.lib
- C:\boost\boost_1_66_0\lib64-msvc-14.1\libboost_program_options-vc141-mt-gd-x64-1_66.lib
@endcode
- Under Build Events > Post-Build Event > Command Line add:
@code
- xcopy /y "C:\mlpack\mlpack-3.1.1\build\Debug\mlpack.dll" $(OutDir)
- xcopy /y "C:\mlpack\mlpack-3.1.1\packages\OpenBLAS.0.2.14.1\lib\native\bin\x64\*.dll" $(OutDir)
- xcopy /y "C:\mlpack\mlpack-3.2.1\build\Debug\mlpack.dll" $(OutDir)
- xcopy /y "C:\mlpack\mlpack-3.2.1\packages\OpenBLAS.0.2.14.1\lib\native\bin\x64\*.dll" $(OutDir)
@endcode
@note Recent versions of Visual Studio set "Conformance Mode" enabled by default. This causes some issues with
doc/tutorials/image/image.txt
+185
View File
@@ -0,0 +1,185 @@
/*!
@file image.txt
@author Mehul Kumar Nirala
@brief Tutorial for how to load and save images in mlpack.
@page imagetutorial Image Utilities tutorial
@section intro_imagetut Introduction
Image datasets are becoming increasingly popular in deep learning.
mlpack's image saving/loading functionality is based on [stb/](https://github.com/nothings/stb).
@section toc_imagetut Table of Contents
This tutorial is split into the following sections:
- \ref intro_imagetut
- \ref toc_imagetut
- \ref model_api_imagetut
- \ref imageinfo_api_imagetut
- \ref load_api_imagetut
- \ref save_api_imagetut
@section model_api_imagetut Model API
Image utilities supports loading and saving of images.
It supports filetypes "jpg", "png", "tga","bmp", "psd", "gif", "hdr", "pic", "pnm" for loading and "jpg", "png", "tga", "bmp", "hdr" for saving.
The datatype associated is unsigned char to support RGB values in the range 1-255. To feed data into the network typecast of `arma::Mat` may be required. Images are stored in matrix as (width * height * channels, NumberOfImages). Therefore imageMatrix.col(0) would be the first image if images are loaded in imageMatrix.
@section imageinfo_api_imagetut ImageInfo
ImageInfo class contains the metadata of the images.
@code
/**
* Instantiate the ImageInfo object with the image width, height, channels.
*
* @param width Image width.
* @param height Image height.
* @param channels number of channels in the image.
*/
ImageInfo(const size_t width,
const size_t height,
const size_t channels);
@endcode
Other public memebers include:
- flipVertical Flip the image vertical upon loading.
- quality Compression of the image if saved as jpg (0-100).
@section load_api_imagetut Load
Standalone loading of images.
@code
/**
* Load the image file into the given matrix.
*
* @param filename Name of the image file.
* @param matrix Matrix to load the image into.
* @param info An object of ImageInfo class.
* @param fatal If an error should be reported as fatal (default false).
* @param transpose If true, transpose the matrix after loading.
* @return Boolean value indicating success or failure of load.
*/
template<typename eT>
bool Load(const std::string& filename,
arma::Mat<eT>& matrix,
ImageInfo& info,
const bool fatal,
const bool transpose);
@endcode
Loading a test image. It also fills up the ImageInfo class object.
@code
data::ImageInfo info;
data::Load("test_image.png", matrix, info, false, true);
@endcode
ImageInfo requires height, width, number of channels of the image.
@code
size_t height = 64, width = 64, channels = 1;
data::ImageInfo info(width, height, channels);
@endcode
More than one image can be loaded into the same matrix.
Loading multiple images:
@code
/**
* Load the image file into the given matrix.
*
* @param files A vector consisting of filenames.
* @param matrix Matrix to save the image from.
* @param info An object of ImageInfo class.
* @param fatal If an error should be reported as fatal (default false).
* @param transpose If true, transpose the matrix after loading.
* @return Boolean value indicating success or failure of load.
*/
template<typename eT>
bool Load(const std::vector<std::string>& files,
arma::Mat<eT>& matrix,
ImageInfo& info,
const bool fatal,
const bool transpose);
@endcode
@code
data::ImageInfo info;
std::vector<std::string>> files{"test_image1.bmp","test_image2.bmp"};
data::load(files, matrix, info, false, true);
@endcode
@section save_api_imagetut Save
Save images expects a matrix of type unsigned char in the form (width * height * channels, NumberOfImages).
Just like load it can be used to save one image or multiple images. Besides image data it also expects the shape of the image as input (width, height, channels).
Saving one image:
@code
/**
* Save the image file from the given matrix.
*
* @param filename Name of the image file.
* @param matrix Matrix to save the image from.
* @param info An object of ImageInfo class.
* @param fatal If an error should be reported as fatal (default false).
* @param transpose If true, transpose the matrix after loading.
* @return Boolean value indicating success or failure of load.
*/
template<typename eT>
bool Save(const std::string& filename,
arma::Mat<eT>& matrix,
ImageInfo& info,
const bool fatal,
const bool transpose);
@endcode
@code
data::ImageInfo info;
info.width = info.height = 25;
info.channels = 3;
info.quality = 90;
data::Save("test_image.bmp", matrix, info, false, true);
@endcode
If the matrix contains more than one image, only the first one is saved.
Saving multiple images:
@code
/**
* Save the image file from the given matrix.
*
* @param files A vector consisting of filenames.
* @param matrix Matrix to save the image from.
* @param info An object of ImageInfo class.
* @param fatal If an error should be reported as fatal (default false).
* @param transpose If true, transpose the matrix after loading.
* @return Boolean value indicating success or failure of load.
*/
template<typename eT>
bool Save(const std::vector<std::string>& files,
arma::Mat<eT>& matrix,
ImageInfo& info,
const bool fatal,
const bool transpose);
@endcode
@code
data::ImageInfo info;
info.width = info.height = 25;
info.channels = 3;
info.quality = 90;
std::vector<std::string>> files{"test_image1.bmp", "test_image2.bmp"};
data::Save(files, matrix, info, false, true);
@endcode
Multiple images are saved according to the vector of filenames specified.
*/
src/mlpack/CMakeLists.txt
+1 -1
View File
@@ -44,7 +44,7 @@ target_link_libraries(mlpack ${MLPACK_LIBRARIES})
set_target_properties(mlpack
PROPERTIES
VERSION 3.1
VERSION 3.2
SOVERSION 3
)
src/mlpack/bindings/python/PythonInstall.cmake
+1 -1
View File
@@ -5,7 +5,7 @@
if (DEFINED ENV{DESTDIR})
execute_process(COMMAND ${PYTHON_EXECUTABLE}
"${CMAKE_BINARY_DIR}/src/mlpack/bindings/python/setup.py" install
--prefix=${CMAKE_INSTALL_PREFIX} --root="$ENV{DESTDIR}"
--prefix=${CMAKE_INSTALL_PREFIX} --root=$ENV{DESTDIR}
WORKING_DIRECTORY "${CMAKE_BINARY_DIR}/src/mlpack/bindings/python/"
RESULT_VARIABLE setup_res)
else ()
src/mlpack/bindings/python/setup.py.in
+3 -1
View File
@@ -108,5 +108,7 @@ setup(name='mlpack',
cmdclass={ 'build_ext': build_ext },
ext_modules = modules,
setup_requires=['cython', 'pytest-runner'],
tests_require=['pytest'],
tests_require=['pytest>3;python_version>"3.4"', 'pytest>3,<=4.6;python_version<="3.4"',
'more-itertools>=4.0.0,<6.0.0;python_version<="2.7"',
'more-itertools>=4.0.0;python_version>"2.7"'],
zip_safe = False)
src/mlpack/core/data/CMakeLists.txt
+2
View File
@@ -10,6 +10,7 @@ set(SOURCES
load_csv.hpp
load_csv.cpp
load.hpp
load_image_impl.hpp
load_model_impl.hpp
load_vec_impl.hpp
load_impl.hpp
@@ -38,6 +39,7 @@ endforeach()
# Add subdirectories.
add_subdirectory(imputation_methods)
add_subdirectory(map_policies)
add_subdirectory(scaler_methods)
# Append sources (with directory name) to list of all mlpack sources (used at
# parent scope).
src/mlpack/core/data/image_info.hpp
+110
View File
@@ -0,0 +1,110 @@
/**
* @file image_info.hpp
* @author Mehul Kumar Nirala
*
* An image information holder.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_CORE_DATA_IMAGE_INFO_HPP
#define MLPACK_CORE_DATA_IMAGE_INFO_HPP
#include <mlpack/prereqs.hpp>
#include "extension.hpp"
#ifdef HAS_STB // Compile this only if stb is present.
#define STB_IMAGE_STATIC
#define STB_IMAGE_IMPLEMENTATION
#include <stb_image.h>
#define STB_IMAGE_WRITE_STATIC
#define STB_IMAGE_WRITE_IMPLEMENTATION
#include <stb_image_write.h>
#endif
namespace mlpack {
namespace data {
#ifdef HAS_STB // Compile this only if stb is present.
/**
* Checks if the given image filename is supported.
*
* @param filename Name of the image file.
* @return Boolean value indicating success if it is an image.
*/
inline bool ImageFormatSupported(const std::string& fileName,
const bool save = false);
#endif
/**
* Implements meta-data of images required by data::Load and
* data::Save for loading and saving images into arma::Mat.
*/
class ImageInfo
{
public:
/**
* Instantiate the ImageInfo object with the given image width, height,
* number of channels and quality parameter.
*
* @param width Image width.
* @param height Image height.
* @param channels Number of channels in the image.
* @param quality Compression of the image if saved as jpg (0 - 100).
*/
ImageInfo(const size_t width = 0,
const size_t height = 0,
const size_t channels = 3,
const size_t quality = 90);
//! Get the image width.
const size_t& Width() const { return width; }
//! Modify the image width.
size_t& Width() { return width; }
//! Get the image height.
const size_t& Height() const { return height; }
//! Modify the image height.
size_t& Height() { return height; }
//! Get the image channels.
const size_t& Channels() const { return channels; }
//! Modify the image channels.
size_t& Channels() { return channels; }
//! Get the image quality.
const size_t& Quality() const { return quality; }
//! Modify the image quality.
size_t& Quality() { return quality; }
private:
// To store the image width.
size_t width;
// To store the image height.
size_t height;
// To store the number of channels in the image.
size_t channels;
// Compression of the image if saved as jpg (0 - 100).
size_t quality;
};
} // namespace data
} // namespace mlpack
// Include implementation of Image.
#include "image_info_impl.hpp"
#endif
src/mlpack/core/data/image_info_impl.hpp
+77
View File
@@ -0,0 +1,77 @@
/**
* @file image_info_impl.hpp
* @author Mehul Kumar Nirala
*
* An image information holder implementation.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_CORE_DATA_IMAGE_INFO_IMPL_HPP
#define MLPACK_CORE_DATA_IMAGE_INFO_IMPL_HPP
#ifdef HAS_STB // Compile this only if stb is present.
// In case it hasn't been included yet.
#include "image_info.hpp"
namespace mlpack {
namespace data {
static const std::vector<std::string> loadFileTypes({"jpg", "png", "tga",
"bmp", "psd", "gif", "hdr", "pic", "pnm", "jpeg"});
static const std::vector<std::string> saveFileTypes({"jpg", "png", "tga",
"bmp", "hdr"});
inline bool ImageFormatSupported(const std::string& fileName, const bool save)
{
if (save)
{
// Iterate over all supported file types that can be saved.
for (auto extension : saveFileTypes)
{
if (extension == Extension(fileName))
return true;
}
}
else
{
// Iterate over all supported file types that can be loaded.
for (auto extension : loadFileTypes)
{
if (extension == Extension(fileName))
return true;
}
}
return false;
}
} // namespace data
} // namespace mlpack
#endif // HAS_STB.
namespace mlpack {
namespace data {
inline ImageInfo::ImageInfo(const size_t width,
const size_t height,
const size_t channels,
const size_t quality) :
width(width),
height(height),
channels(channels),
quality(quality)
{
// Do nothing.
}
} // namespace data
} // namespace mlpack
#endif
src/mlpack/core/data/load.hpp
+44
View File
@@ -20,6 +20,7 @@
#include "format.hpp"
#include "dataset_mapper.hpp"
#include "image_info.hpp"
namespace mlpack {
namespace data /** Functions to load and save matrices and models. */ {
@@ -287,6 +288,47 @@ bool Load(const std::string& filename,
const bool fatal = false,
format f = format::autodetect);
/**
* Image load/save interfaces.
*/
#ifdef HAS_STB
/**
* Load the image file into the given matrix.
*
* @param filename Name of the image file.
* @param matrix Matrix to load the image into.
* @param info An object of ImageInfo class.
* @param fatal If an error should be reported as fatal (default false).
* @param transpose If true, transpose the matrix after loading.
* @return Boolean value indicating success or failure of load.
*/
template<typename eT>
bool Load(const std::string& filename,
arma::Mat<eT>& matrix,
ImageInfo& info,
const bool fatal = false,
const bool transpose = true);
/**
* Load the image file into the given matrix.
*
* @param files A vector consisting of filenames.
* @param matrix Matrix to save the image from.
* @param info An object of ImageInfo class.
* @param fatal If an error should be reported as fatal (default false).
* @param transpose If true, transpose the matrix after loading.
* @return Boolean value indicating success or failure of load.
*/
template<typename eT>
bool Load(const std::vector<std::string>& files,
arma::Mat<eT>& matrix,
ImageInfo& info,
const bool fatal = false,
const bool transpose = true);
#endif // HAS_STB.
} // namespace data
} // namespace mlpack
@@ -294,5 +336,7 @@ bool Load(const std::string& filename,
#include "load_model_impl.hpp"
// Include implementation of Load() for vectors.
#include "load_vec_impl.hpp"
// Include implementation of Load() for images.
#include "load_image_impl.hpp"
#endif
src/mlpack/core/data/load_image_impl.hpp
+148
View File
@@ -0,0 +1,148 @@
/**
* @file load_image_impl.hpp
* @author Mehul Kumar Nirala
*
* An image loading utility implementation.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_CORE_DATA_LOAD_IMAGE_IMPL_HPP
#define MLPACK_CORE_DATA_LOAD_IMAGE_IMPL_HPP
// In case it hasn't been included yet.
#include "load.hpp"
namespace mlpack {
namespace data {
#ifdef HAS_STB // Compile this only if stb is present.
// Image loading API.
template<typename eT>
bool Load(const std::string& filename,
arma::Mat<eT>& matrix,
ImageInfo& info,
const bool /* fatal */,
const bool transpose)
{
Timer::Start("loading_image");
unsigned char* image;
if (!ImageFormatSupported(filename))
{
std::ostringstream oss;
oss << "File type " << Extension(filename) << " not supported.\n";
oss << "Currently it supports ";
for (auto extension : loadFileTypes)
oss << " " << extension;
oss << std::endl;
throw std::runtime_error(oss.str());
return false;
}
stbi_set_flip_vertically_on_load(transpose);
// Temporary variables needed as stb_image.h supports int parameters.
int tempWidth, tempHeight, tempChannels;
// For grayscale images.
if (info.Channels() == 1)
{
image = stbi_load(filename.c_str(), &tempWidth, &tempHeight, &tempChannels,
STBI_grey);
}
else
{
image = stbi_load(filename.c_str(), &tempWidth, &tempHeight, &tempChannels,
STBI_rgb);
}
if (tempWidth <= 0 || tempHeight <= 0)
{
std::ostringstream oss;
oss << "Image '" << filename << "' not found." << std::endl;
free(image);
throw std::runtime_error(oss.str());
return false;
}
info.Width() = tempWidth;
info.Height() = tempHeight;
info.Channels() = tempChannels;
// Copy image into armadillo Mat.
matrix = arma::Mat<unsigned char>(image, info.Width() * info.Height() *
info.Channels(), 1, true, true);
// Free the image pointer.
free(image);
Timer::Stop("loading_image");
return true;
}
// Image loading API for multiple files.
template<typename eT>
bool Load(const std::vector<std::string>& files,
arma::Mat<eT>& matrix,
ImageInfo& info,
const bool fatal,
const bool transpose)
{
if (files.size() == 0)
{
std::ostringstream oss;
oss << "Files vector is empty." << std::endl;
throw std::runtime_error(oss.str());
return false;
}
arma::Mat<unsigned char> img;
bool status = Load(files[0], img, info, fatal, transpose);
// Decide matrix dimension using the image height and width.
matrix.set_size(info.Width() * info.Height() * info.Channels(), files.size());
matrix.col(0) = img;
for (size_t i = 1; i < files.size() ; i++)
{
arma::Mat<unsigned char> colImg(matrix.colptr(i), matrix.n_rows, 1,
false, true);
status &= Load(files[i], colImg, info, fatal, transpose);
}
return status;
}
#else // No STB.
template<typename eT>
bool Load(const std::string& filename,
arma::Mat<eT>& matrix,
ImageInfo& info,
const bool fatal = false,
const bool transpose = true)
{
throw std::runtime_error("Load(): HAS_STB is not defined, "
"so STB is not available and images cannot be loaded!");
}
template<typename eT>
bool Load(const std::vector<std::string>& files,
arma::Mat<eT>& matrix,
ImageInfo& info,
const bool fatal = false,
const bool transpose = true)
{
throw std::runtime_error("Load(): HAS_STB is not defined, "
"so STB is not available and images cannot be loaded!");
}
#endif // HAS_STB.
} // namespace data
} // namespace mlpack
#endif
src/mlpack/core/data/save.hpp
+39
View File
@@ -19,6 +19,7 @@
#include <string>
#include "format.hpp"
#include "image_info.hpp"
namespace mlpack {
namespace data /** Functions to load and save matrices. */ {
@@ -90,6 +91,44 @@ bool Save(const std::string& filename,
const bool fatal = false,
format f = format::autodetect);
#ifdef HAS_STB
/**
* Save the image file from the given matrix.
*
* @param filename Name of the image file.
* @param matrix Matrix to save the image from.
* @param info An object of ImageInfo class.
* @param fatal If an error should be reported as fatal (default false).
* @param transpose If true, transpose the matrix after loading.
* @return Boolean value indicating success or failure of load.
*/
template<typename eT>
bool Save(const std::string& filename,
arma::Mat<eT>& matrix,
ImageInfo& info,
const bool fatal = false,
const bool transpose = true);
/**
* Save the image file from the given matrix.
*
* @param files A vector consisting of filenames.
* @param matrix Matrix to save the image from.
* @param info An object of ImageInfo class.
* @param fatal If an error should be reported as fatal (default false).
* @param transpose If true, transpose the matrix after loading.
* @return Boolean value indicating success or failure of load.
*/
template<typename eT>
bool Save(const std::vector<std::string>& files,
arma::Mat<eT>& matrix,
ImageInfo& info,
const bool fatal = false,
const bool transpose = true);
#endif // HAS_STB.
} // namespace data
} // namespace mlpack
src/mlpack/core/data/save_impl.hpp
+141
View File
@@ -283,6 +283,147 @@ bool Save(const std::string& filename,
}
}
#ifdef HAS_STB
// Image saving API.
template<typename eT>
bool Save(const std::string& filename,
arma::Mat<eT>& matrix,
ImageInfo& info,
const bool fatal,
const bool transpose)
{
Timer::Start("saving_image");
// We transpose by default. So, un-transpose if necessary.
if (!transpose)
matrix = arma::trans(matrix);
int tempWidth, tempHeight, tempChannels, tempQuality;
tempWidth = info.Width();
tempHeight = info.Height();
tempChannels = info.Channels();
tempQuality = info.Quality();
if (!ImageFormatSupported(filename, true))
{
std::ostringstream oss;
oss << "File type " << Extension(filename) << " not supported.\n";
oss << "Currently it supports ";
for (auto extension : saveFileTypes)
oss << ", " << extension;
oss << std::endl;
throw std::runtime_error(oss.str());
return false;
}
if (matrix.n_cols > 1)
{
std::cout << "Input Matrix contains more than 1 image." << std::endl;
std::cout << "Only the firstimage will be saved!" << std::endl;
}
stbi_flip_vertically_on_write(transpose);
bool status = false;
try
{
unsigned char* image = matrix.memptr();
if ("png" == Extension(filename))
{
status = stbi_write_png(filename.c_str(), tempWidth, tempHeight,
tempChannels, image, tempWidth * tempChannels);
}
else if ("bmp" == Extension(filename))
{
status = stbi_write_bmp(filename.c_str(), tempWidth, tempHeight,
tempChannels, image);
}
else if ("tga" == Extension(filename))
{
status = stbi_write_tga(filename.c_str(), tempWidth, tempHeight,
tempChannels, image);
}
else if ("hdr" == Extension(filename))
{
status = stbi_write_hdr(filename.c_str(), tempWidth, tempHeight,
tempChannels, reinterpret_cast<float*>(image));
}
else if ("jpg" == Extension(filename))
{
status = stbi_write_jpg(filename.c_str(), tempWidth, tempHeight,
tempChannels, image, tempQuality);
}
}
catch (std::exception& e)
{
Timer::Stop("saving_image");
if (fatal)
Log::Fatal << e.what() << std::endl;
Log::Warn << e.what() << std::endl;
return false;
}
Timer::Stop("saving_image");
return status;
}
// Image saving API for multiple files.
template<typename eT>
bool Save(const std::vector<std::string>& files,
arma::Mat<eT>& matrix,
ImageInfo& info,
const bool fatal,
const bool transpose)
{
if (files.size() == 0)
{
std::ostringstream oss;
oss << "Files vector is empty." << std::endl;
throw std::runtime_error(oss.str());
return false;
}
// We transpose by default. So, un-transpose if necessary.
if (!transpose)
matrix = arma::trans(matrix);
arma::Mat<unsigned char> img;
bool status = Save(files[0], img, info, fatal, transpose);
// Decide matrix dimension using the image height and width.
matrix.set_size(info.Width() * info.Height() * info.Channels(), files.size());
matrix.col(0) = img;
for (size_t i = 1; i < files.size() ; i++)
{
arma::Mat<unsigned char> colImg(matrix.colptr(i), matrix.n_rows, 1,
false, true);
status &= Save(files[i], colImg, info, fatal, transpose);
}
return status;
}
#else
template<typename eT>
bool Save(const std::string& filename,
arma::Mat<eT>& matrix,
ImageInfo& info,
const bool fatal = false,
const bool transpose = true)
{
throw std::runtime_error("Save(): HAS_STB is not defined, "
"so STB is not available and images cannot be saved!");
}
template<typename eT>
bool Save(const std::vector<std::string>& files,
arma::Mat<eT>& matrix,
ImageInfo& info,
const bool fatal = false,
const bool transpose = true)
{
throw std::runtime_error("Save(): HAS_STB is not defined, "
"so STB is not available and images cannot be saved!");
}
#endif // HAS_STB.
} // namespace data
} // namespace mlpack
src/mlpack/core/data/scaler_methods/CMakeLists.txt
+19
View File
@@ -0,0 +1,19 @@
# Define the files we need to compile
# Anything not in this list will not be compiled into mlpack.
set(SOURCES
min_max_scaler.hpp
max_abs_scaler.hpp
standard_scaler.hpp
mean_normalization.hpp
pca_whitening.hpp
zca_whitening.hpp
)
# Add directory name to sources.
set(DIR_SRCS)
foreach(file ${SOURCES})
set(DIR_SRCS ${DIR_SRCS} ${CMAKE_CURRENT_SOURCE_DIR}/${file})
endforeach()
# Append sources (with directory name) to list of all mlpack sources (used at
# the parent scope).
set(MLPACK_SRCS ${MLPACK_SRCS} ${DIR_SRCS} PARENT_SCOPE)
src/mlpack/core/data/scaler_methods/max_abs_scaler.hpp
+122
View File
@@ -0,0 +1,122 @@
/**
* @file max_abs_scaler.hpp
* @author Jeffin Sam
*
* MaxAbsScaler class to scale features.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_CORE_DATA_MAX_ABS_SCALE_HPP
#define MLPACK_CORE_DATA_MAX_ABS_SCALE_HPP
#include <mlpack/prereqs.hpp>
namespace mlpack {
namespace data {
/**
* A simple MaxAbs Scaler class.
*
* Given an input dataset this class helps you to scale each
* feature by its maximum absolute value.
*
* \[z = x / max(abs(x))\]
*
* where max(abs(x)) is maximum absolute value of feature.
*
* @code
* arma::mat input;
* Load("train.csv", input);
* arma::mat output;
*
* // Fit the features.
* MaxAbsScaler scale;
* scale.Fit(input)
*
* // Scale the features.
* scale.Transform(input, output);
*
* // Retransform the input.
* scale.InverseTransform(output, input);
* @endcode
*/
class MaxAbsScaler
{
public:
/**
* Function to fit features, to find out the min max and scale.
*
* @param input Dataset to fit.
*/
template<typename MatType>
void Fit(const MatType& input)
{
itemMin = arma::min(input, 1);
itemMax = arma::max(input, 1);
scale = arma::max(arma::abs(itemMin), arma::abs(itemMax));
// Handling zeros in scale vector.
scale.for_each([](arma::vec::elem_type& val) { val =
(val == 0) ? 1 : val; });
}
/**
* Function to scale features.
*
* @param input Dataset to scale features.
* @param output Output matrix with scaled features.
*/
template<typename MatType>
void Transform(const MatType& input, MatType& output)
{
if (scale.is_empty())
{
throw std::runtime_error("Call Fit() before Transform(), please"
" refer to the documentation.");
}
output.copy_size(input);
output = input.each_col() / scale;
}
/**
* Function to retrieve original dataset.
*
* @param input Scaled dataset.
* @param output Output matrix with original Dataset.
*/
template<typename MatType>
void InverseTransform(const MatType& input, MatType& output)
{
output.copy_size(input);
output = input.each_col() % scale;
}
//! Get the Min row vector.
const arma::vec& ItemMin() const { return itemMin; }
//! Get the Max row vector.
const arma::vec& ItemMax() const { return itemMax; }
//! Get the Scale row vector.
const arma::vec& Scale() const { return scale; }
template<typename Archive>
void serialize(Archive& ar, const unsigned int /* version */)
{
ar & BOOST_SERIALIZATION_NVP(itemMin);
ar & BOOST_SERIALIZATION_NVP(itemMax);
ar & BOOST_SERIALIZATION_NVP(scale);
}
private:
// Vector which holds minimum of each feature.
arma::vec itemMin;
// Vector which holds maximum of each feature.
arma::vec itemMax;
// Vector which is used to scale up each feature.
arma::vec scale;
}; // class MaxAbsScaler
} // namespace data
} // namespace mlpack
#endif
src/mlpack/core/data/scaler_methods/mean_normalization.hpp
+129
View File
@@ -0,0 +1,129 @@
/**
* @file mean_normalization.hpp
* @author Jeffin Sam
*
* MeanNormalization class to scale features.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_CORE_DATA_MEAN_NORMALIZATION_HPP
#define MLPACK_CORE_DATA_MEAN_NORMALIZATION_HPP
#include <mlpack/prereqs.hpp>
namespace mlpack {
namespace data {
/**
* A simple Mean Normalization class
*
* Given an input dataset this class helps you to normalize each
* feature.
*
* \[z = x - average(x) / (max(x) - min(x))\]
*
* where x is an original value,z is the normalized value.
*
* @code
* arma::mat input;
* Load("train.csv", input);
* arma::mat output;
*
* // Fit the features.
* MeanNormalization scale;
* scale.Fit(input)
*
* // Scale the features.
* scale.Transform(input, output);
*
* // Retransform the input.
* scale.InverseTransform(output, input);
* @endcode
*/
class MeanNormalization
{
public:
/**
* Function to fit features, to find out the min max and scale.
*
* @param input Dataset to fit.
*/
template<typename MatType>
void Fit(const MatType& input)
{
itemMean = arma::mean(input, 1);
itemMin = arma::min(input, 1);
itemMax = arma::max(input, 1);
scale = itemMax - itemMin;
// Handling zeros in scale vector.
scale.for_each([](arma::vec::elem_type& val) { val =
(val == 0) ? 1 : val; });
}
/**
* Function to scale features.
*
* @param input Dataset to scale features.
* @param output Output matrix with scaled features.
*/
template<typename MatType>
void Transform(const MatType& input, MatType& output)
{
if (itemMean.is_empty() || scale.is_empty())
{
throw std::runtime_error("Call Fit() before Transform(), please"
" refer to the documentation.");
}
output.copy_size(input);
output = (input.each_col() - itemMean).each_col() / scale;
}
/**
* Function to retrieve original dataset.
*
* @param input Scaled dataset.
* @param output Output matrix with original Dataset.
*/
template<typename MatType>
void InverseTransform(const MatType& input, MatType& output)
{
output.copy_size(input);
output = (input.each_col() % scale).each_col() + itemMean;
}
//! Get the Mean row vector.
const arma::vec& ItemMean() const { return itemMean; }
//! Get the Min row vector.
const arma::vec& ItemMin() const { return itemMin; }
//! Get the Max row vector.
const arma::vec& ItemMax() const { return itemMax; }
//! Get the Scale row vector.
const arma::vec& Scale() const { return scale; }
template<typename Archive>
void serialize(Archive& ar, const unsigned int /* version */)
{
ar & BOOST_SERIALIZATION_NVP(itemMin);
ar & BOOST_SERIALIZATION_NVP(itemMax);
ar & BOOST_SERIALIZATION_NVP(scale);
ar & BOOST_SERIALIZATION_NVP(itemMean);
}
private:
// Vector which holds mean of each feature.
arma::vec itemMean;
// Vector which holds minimum of each feature.
arma::vec itemMin;
// Vector which holds maximum of each feature.
arma::vec itemMax;
// Vector which is used to scale up each feature.
arma::vec scale;
}; // class MeanNormalization
} // namespace data
} // namespace mlpack
#endif
src/mlpack/core/data/scaler_methods/min_max_scaler.hpp
+159
View File
@@ -0,0 +1,159 @@
/**
* @file min_max_scaler.hpp
* @author Jeffin Sam
*
* MinMaxScaler class to scale features.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_CORE_DATA_SCALE_HPP
#define MLPACK_CORE_DATA_SCALE_HPP
#include <mlpack/prereqs.hpp>
namespace mlpack {
namespace data {
/**
* A simple MinMax Scaler class
*
* Given an input dataset this class helps you to scale each
* feature to a given range.
*
* \[z = scale * x + scaleMin - min(x) * scale\]
* \[scale = (scaleMax - scaleMin) / (max(x) - min(x))\]
*
* where scaleMin, scaleMax = feature_range and min(x), max(x)
* are the minimum and maximum value of x respectively.
*
* @code
* arma::mat input;
* Load("train.csv", input);
* arma::mat output;
*
* // Fit the features.
* MinMaxScaler scale;
* scale.Fit(input)
*
* // Scale the features.
* scale.Transform(input, output);
*
* // Retransform the input.
* scale.InverseTransform(output, input);
* @endcode
*/
class MinMaxScaler
{
public:
/**
* Default constructor
*
* @param min Lower range of scaling.
* @param max Upper range of scaling.
*/
MinMaxScaler(const double min = 0, const double max = 1)
{
scaleMin = min;
scaleMax = max;
// Ensure scaleMin is smaller than scaleMax.
if (scaleMin > scaleMax)
{
throw std::runtime_error("Range is not appropriate");
}
}
/**
* Function to fit features, to find out the min max and scale.
*
* @param input Dataset to fit.
*/
template<typename MatType>
void Fit(const MatType& input)
{
itemMin = arma::min(input, 1);
itemMax = arma::max(input, 1);
scale = itemMax - itemMin;
// Handle zeros in scale vector.
scale.for_each([](arma::vec::elem_type& val) { val =
(val == 0) ? 1 : val; });
scale = (scaleMax - scaleMin) / scale;
scalerowmin.copy_size(itemMin);
scalerowmin.fill(scaleMin);
scalerowmin = scalerowmin - itemMin % scale;
}
/**
* Function to scale features.
*
* @param input Dataset to scale features.
* @param output Output matrix with scaled features.
*/
template<typename MatType>
void Transform(const MatType& input, MatType& output)
{
if (scalerowmin.is_empty() || scale.is_empty())
{
throw std::runtime_error("Call Fit() before Transform(), please"
" refer to the documentation.");
}
output.copy_size(input);
output = (input.each_col() % scale).each_col() + scalerowmin;
}
/**
* Function to retrieve original dataset.
*
* @param input Scaled dataset.
* @param output Output matrix with original Dataset.
*/
template<typename MatType>
void InverseTransform(const MatType& input, MatType& output)
{
output.copy_size(input);
output = (input.each_col() - scalerowmin).each_col() / scale;
}
//! Get the Min row vector.
const arma::vec& ItemMin() const { return itemMin; }
//! Get the Max row vector.
const arma::vec& ItemMax() const { return itemMax; }
//! Get the Scale row vector.
const arma::vec& Scale() const { return scale; }
//! Get the upper range parameter.
double ScaleMax() const { return scaleMax; }
//! Get the lower range parameter.
double ScaleMin() const { return scaleMin; }
template<typename Archive>
void serialize(Archive& ar, const unsigned int /* version */)
{
ar & BOOST_SERIALIZATION_NVP(itemMin);
ar & BOOST_SERIALIZATION_NVP(itemMax);
ar & BOOST_SERIALIZATION_NVP(scale);
ar & BOOST_SERIALIZATION_NVP(scaleMin);
ar & BOOST_SERIALIZATION_NVP(scaleMax);
ar & BOOST_SERIALIZATION_NVP(scalerowmin);
}
private:
// Vector which holds minimum of each feature.
arma::vec itemMin;
// Vector which holds maximum of each feature.
arma::vec itemMax;
// Scale vector which is used to scale up each feature.
arma::vec scale;
// Lower value for range.
double scaleMin;
// Upper value for range.
double scaleMax;
// Column vector of scalemin
arma::vec scalerowmin;
}; // class MinMaxScaler
} // namespace data
} // namespace mlpack
#endif
src/mlpack/core/data/scaler_methods/pca_whitening.hpp
+145
View File
@@ -0,0 +1,145 @@
/**
* @file pca_whitening.hpp
* @author Jeffin Sam
*
* Whitening scaling to scale features, Using PCA Whitening.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_CORE_DATA_PCA_WHITENING_SCALE_HPP
#define MLPACK_CORE_DATA_PCA_WHITENING_SCALE_HPP
#include <mlpack/prereqs.hpp>
#include <mlpack/core/math/lin_alg.hpp>
namespace mlpack {
namespace data {
/**
* A simple PCAWhitening class.
*
* Whitens a matrix using the eigendecomposition of the covariance matrix.
* Whitening means the covariance matrix of the result is the identity matrix.
*
* For whitening related formula and more info, check the link below.
* http://ufldl.stanford.edu/tutorial/unsupervised/PCAWhitening/
*
* @code
* arma::mat input;
* Load("train.csv", input);
* arma::mat output;
*
* // Fit the features.
* PCAWhitening scale;
* scale.Fit(input)
*
* // Scale the features.
* scale.Transform(input, output);
*
* // Retransform the input.
* scale.InverseTransform(output, input);
* @endcode
*/
class PCAWhitening
{
public:
/**
* A constructor to set the regularization parameter.
*
* @param eps Regularization parameter.
*/
PCAWhitening(double eps = 0.00005)
{
epsilon = eps;
// Ensure scaleMin is smaller than scaleMax.
if (epsilon < 0)
{
throw std::runtime_error("Regularization parameter is not correct");
}
}
/**
* Function to fit features, to find out the min max and scale.
*
* @param input Dataset to fit.
*/
template<typename MatType>
void Fit(const MatType& input)
{
itemMean = arma::mean(input, 1);
// Get eigenvectors and eigenvalues of covariance of input matrix.
eig_sym(eigenValues, eigenVectors, mlpack::math::ColumnCovariance(
input.each_col() - itemMean));
eigenValues += epsilon;
}
/**
* Function for PCA whitening.
*
* @param input Dataset to scale features.
* @param output Output matrix with whitened features.
*/
template<typename MatType>
void Transform(const MatType& input, MatType& output)
{
if (eigenValues.is_empty() || eigenVectors.is_empty())
{
throw std::runtime_error("Call Fit() before Transform(), please"
" refer to the documentation.");
}
output.copy_size(input);
output = (input.each_col() - itemMean);
output = arma::diagmat(1.0 / (arma::sqrt(eigenValues))) * eigenVectors.t()
* output;
}
/**
* Function to retrieve original dataset.
*
* @param input Scaled dataset.
* @param output Output matrix with original Dataset.
*/
template<typename MatType>
void InverseTransform(const MatType& input, MatType& output)
{
output = arma::diagmat(arma::sqrt(eigenValues)) * inv(eigenVectors.t())
* input;
output = (output.each_col() + itemMean);
}
//! Get the mean row vector.
const arma::vec& ItemMean() const { return itemMean; }
//! Get the eigenvalues vector.
const arma::vec& EigenValues() const { return eigenValues; }
//! Get the eigenvector.
const arma::mat& EigenVectors() const { return eigenVectors; }
//! Get the regularization parameter.
const double& Epsilon() const { return epsilon; }
template<typename Archive>
void serialize(Archive& ar, const unsigned int /* version */)
{
ar & BOOST_SERIALIZATION_NVP(eigenValues);
ar & BOOST_SERIALIZATION_NVP(eigenVectors);
ar & BOOST_SERIALIZATION_NVP(itemMean);
ar & BOOST_SERIALIZATION_NVP(epsilon);
}
private:
// Vector which holds mean of each feature.
arma::vec itemMean;
// Mat which hold the eigenvectors.
arma::mat eigenVectors;
// Regularization Paramter.
double epsilon;
// Vector which hold the eigenvalues.
arma::vec eigenValues;
}; // class PCAWhitening
} // namespace data
} // namespace mlpack
#endif
src/mlpack/core/data/scaler_methods/standard_scaler.hpp
+118
View File
@@ -0,0 +1,118 @@
/**
* @file standard_scaler.hpp
* @author Jeffin Sam
*
* StandardScaler class to scale features.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_CORE_DATA_STANDARD_SCALE_HPP
#define MLPACK_CORE_DATA_STANDARD_SCALE_HPP
#include <mlpack/prereqs.hpp>
namespace mlpack {
namespace data {
/**
* A simple Standard Scaler class
*
* Given an input dataset this class helps you to Standardize features
* by removing the mean and scaling to unit variance.
*
* \[z = (x - u) / s\]
*
* where u is the mean of the training samples and s is the standard deviation
* of the training samples.
*
* @code
* arma::mat input;
* Load("train.csv", input);
* arma::mat output;
*
* // Fit the features.
* StandardScaler scale;
* scale.Fit(input)
*
* // Scale the features.
* scale.Transform(input, output);
*
* // Retransform the input.
* scale.InverseTransform(output, input);
* @endcode
*/
class StandardScaler
{
public:
/**
* Function to fit features, to find out the min max and scale.
*
* @param input Dataset to fit.
*/
template<typename MatType>
void Fit(const MatType& input)
{
itemMean = arma::mean(input, 1);
itemStdDev = arma::stddev(input, 1, 1);
// Handle zeros in scale vector.
itemStdDev.for_each([](arma::vec::elem_type& val) { val =
(val == 0) ? 1 : val; });
}
/**
* Function to scale features.
*
* @param input Dataset to scale features.
* @param output Output matrix with scaled features.
*/
template<typename MatType>
void Transform(const MatType& input, MatType& output)
{
if (itemMean.is_empty() || itemStdDev.is_empty())
{
throw std::runtime_error("Call Fit() before Transform(), please"
" refer to the documentation.");
}
output.copy_size(input);
output = (input.each_col() - itemMean).each_col() / itemStdDev;
}
/**
* Function to retrieve original dataset.
*
* @param input Scaled dataset.
* @param output Output matrix with original Dataset.
*/
template<typename MatType>
void InverseTransform(const MatType& input, MatType& output)
{
output.copy_size(input);
output = (input.each_col() % itemStdDev).each_col() + itemMean;
}
//! Get the mean row vector.
const arma::vec& ItemMean() const { return itemMean; }
//! Get the standard deviation row vector.
const arma::vec& ItemStdDev() const { return itemStdDev; }
template<typename Archive>
void serialize(Archive& ar, const unsigned int /* version */)
{
ar & BOOST_SERIALIZATION_NVP(itemMean);
ar & BOOST_SERIALIZATION_NVP(itemStdDev);
}
private:
// Vector which holds mean of each feature.
arma::vec itemMean;
// Vector which holds standard devation of each feature.
arma::vec itemStdDev;
}; // class StandardScaler
} // namespace data
} // namespace mlpack
#endif
src/mlpack/core/data/scaler_methods/zca_whitening.hpp
+121
View File
@@ -0,0 +1,121 @@
/**
* @file zca_whitening.hpp
* @author Jeffin Sam
*
* Whitening scaling to scale features, Using ZCA Whitening.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_CORE_DATA_ZCA_WHITENING_SCALE_HPP
#define MLPACK_CORE_DATA_ZCA_WHITENING_SCALE_HPP
#include <mlpack/prereqs.hpp>
#include <mlpack/core/math/lin_alg.hpp>
#include <mlpack/core/data/scaler_methods/pca_whitening.hpp>
namespace mlpack {
namespace data {
/**
* A simple ZCAWhitening class.
*
* Whitens a matrix using the eigendecomposition of the covariance matrix.
* Whitening means the covariance matrix of the result is the identity matrix.
*
* For whitening related formula and more info, check the link below.
* http://ufldl.stanford.edu/tutorial/unsupervised/PCAWhitening/
*
* @code
* arma::mat input;
* Load("train.csv", input);
* arma::mat output;
*
* // Fit the features.
* ZCAWhitening scale;
* scale.Fit(input)
*
* // Scale the features.
* scale.Transform(input, output);
*
* // Retransform the input.
* scale.InverseTransform(output, input);
* @endcode
*/
class ZCAWhitening
{
public:
/**
* A constructor to set the regularization parameter.
*
* @param eps Regularization parameter.
*/
ZCAWhitening(double eps = 0.00005)
{
pca = new data::PCAWhitening(eps);
}
/**
* Function to fit features, to find out the min max and scale.
*
* @param input Dataset to fit.
*/
template<typename MatType>
void Fit(const MatType& input)
{
pca->Fit(input);
}
/**
* Function for ZCA whitening.
*
* @param input Dataset to scale features.
* @param output Output matrix with whitened features.
*/
template<typename MatType>
void Transform(const MatType& input, MatType& output)
{
pca->Transform(input, output);
output = pca->EigenVectors() * output;
}
/**
* Function to retrieve original dataset.
*
* @param input Scaled dataset.
* @param output Output matrix with original Dataset.
*/
template<typename MatType>
void InverseTransform(const MatType& input, MatType& output)
{
output = inv(pca->EigenVectors()) * arma::diagmat(arma::sqrt(
pca->EigenValues())) * inv(pca->EigenVectors().t()) * input;
output = (output.each_col() + pca->ItemMean());
}
//! Get the mean row vector.
const arma::vec& ItemMean() const { return pca->ItemMean(); }
//! Get the eigenvalues vector.
const arma::vec& EigenValues() const { return pca->EigenValues(); }
//! Get the eigenvector.
const arma::mat& EigenVectors() const { return pca->EigenVectors(); }
//! Get the regularization parameter.
double Epsilon() const { return pca->Epsilon(); }
template<typename Archive>
void serialize(Archive& ar, const unsigned int /* version */)
{
ar & BOOST_SERIALIZATION_NVP(pca);
}
private:
// A pointer to PcaWhitening Class.
PCAWhitening* pca;
}; // class ZCAWhitening
} // namespace data
} // namespace mlpack
#endif
src/mlpack/core/dists/gaussian_distribution.hpp
+6
View File
@@ -168,6 +168,12 @@ class GaussianDistribution
void Covariance(arma::mat&& covariance);
//! Return the invCov.
const arma::mat& InvCov() const { return invCov; }
//! Return the logDetCov.
double LogDetCov() const { return logDetCov; }
/**
* Serialize the distribution.
*/
src/mlpack/core/math/random.hpp
+7
View File
@@ -63,6 +63,13 @@ inline void FixedRandomSeed()
srand((unsigned int) seed);
arma::arma_rng::set_seed(seed);
}
inline void CustomRandomSeed(const size_t seed)
{
randGen.seed((uint32_t) seed);
srand((unsigned int) seed);
arma::arma_rng::set_seed(seed);
}
#endif
/**
src/mlpack/core/metrics/lmetric_impl.hpp
+1 -1
View File
@@ -61,7 +61,7 @@ typename VecTypeA::elem_type LMetric<2, true>::Evaluate(
const VecTypeA& a,
const VecTypeB& b)
{
return sqrt(arma::accu(square(a - b)));
return arma::norm(a - b, 2);
}
template<>
src/mlpack/core/tree/binary_space_tree/dual_tree_traverser_impl.hpp
+12
View File
@@ -56,6 +56,18 @@ DualTreeTraverser<RuleType>::Traverse(
// Store the current traversal info.
traversalInfo = rule.TraversalInfo();
// If both nodes are root nodes, just score them.
if (queryNode.Parent() == NULL && referenceNode.Parent() == NULL)
{
const double rootScore = rule.Score(queryNode, referenceNode);
// If root score is DBL_MAX, don't recurse.
if (rootScore == DBL_MAX)
{
++numPrunes;
return;
}
}
// If both are leaves, we must evaluate the base case.
if (queryNode.IsLeaf() && referenceNode.IsLeaf())
{
src/mlpack/core/tree/binary_space_tree/single_tree_traverser_impl.hpp
+12
View File
@@ -57,6 +57,18 @@ SingleTreeTraverser<RuleType>::Traverse(
}
else
{
// If it's the root node, just score it.
if (referenceNode.Parent() == NULL)
{
const double rootScore = rule.Score(queryIndex, referenceNode);
// If root score is DBL_MAX, don't recurse into that node.
if (rootScore == DBL_MAX)
{
++numPrunes;
return;
}
}
// If either score is DBL_MAX, we do not recurse into that node.
double leftScore = rule.Score(queryIndex, *referenceNode.Left());
double rightScore = rule.Score(queryIndex, *referenceNode.Right());
src/mlpack/core/tree/cover_tree/cover_tree_impl.hpp
+62 -41
View File
@@ -21,6 +21,18 @@
namespace mlpack {
namespace tree {
// Build the statistics, bottom-up.
template<typename TreeType, typename StatisticType>
void BuildStatistics(TreeType* node)
{
// Recurse first.
for (size_t i = 0; i < node->NumChildren(); ++i)
BuildStatistics<TreeType, StatisticType>(&node->Child(i));
// Now build the statistic.
node->Stat() = StatisticType(*node);
}
// Create the cover tree.
template<
typename MetricType,
@@ -88,9 +100,8 @@ CoverTree<MetricType, StatisticType, MatType, RootPointPolicy>::CoverTree(
{
children.push_back(&(old->Child(i)));
// Set its parent correctly, and rebuild the statistic.
// Set its parent correctly.
old->Child(i).Parent() = this;
old->Child(i).Stat() = StatisticType(old->Child(i));
}
// Remove all the children so they don't get erased.
@@ -104,14 +115,19 @@ CoverTree<MetricType, StatisticType, MatType, RootPointPolicy>::CoverTree(
}
// Use the furthest descendant distance to determine the scale of the root
// node.
if (furthestDescendantDistance == 0.0)
// node. Note that if the root is a leaf, we can have scale INT_MIN, but if
// it *isn't* a leaf, we need to mark the scale as one higher than INT_MIN, so
// that the recursions don't fail.
if (furthestDescendantDistance == 0.0 && dataset.n_cols == 1)
scale = INT_MIN;
else if (furthestDescendantDistance == 0.0)
scale = INT_MIN + 1;
else
scale = (int) ceil(log(furthestDescendantDistance) / log(base));
// Initialize statistic.
stat = StatisticType(*this);
// Initialize statistics recursively after the entire tree construction is
// complete.
BuildStatistics<CoverTree, StatisticType>(this);
Log::Info << distanceComps << " distance computations during tree "
<< "construction." << std::endl;
@@ -181,8 +197,6 @@ CoverTree<MetricType, StatisticType, MatType, RootPointPolicy>::CoverTree(
// Set its parent correctly.
old->Child(i).Parent() = this;
// Rebuild the statistic.
old->Child(i).Stat() = StatisticType(old->Child(i));
}
// Remove all the children so they don't get erased.
@@ -196,14 +210,19 @@ CoverTree<MetricType, StatisticType, MatType, RootPointPolicy>::CoverTree(
}
// Use the furthest descendant distance to determine the scale of the root
// node.
if (furthestDescendantDistance == 0.0)
// node. Note that if the root is a leaf, we can have scale INT_MIN, but if
// it *isn't* a leaf, we need to mark the scale as one higher than INT_MIN, so
// that the recursions don't fail.
if (furthestDescendantDistance == 0.0 && dataset.n_cols == 1)
scale = INT_MIN;
else if (furthestDescendantDistance == 0.0)
scale = INT_MIN + 1;
else
scale = (int) ceil(log(furthestDescendantDistance) / log(base));
// Initialize statistic.
stat = StatisticType(*this);
// Initialize statistics recursively after the entire tree construction is
// complete.
BuildStatistics<CoverTree, StatisticType>(this);
Log::Info << distanceComps << " distance computations during tree "
<< "construction." << std::endl;
@@ -272,9 +291,8 @@ CoverTree<MetricType, StatisticType, MatType, RootPointPolicy>::CoverTree(
{
children.push_back(&(old->Child(i)));
// Set its parent correctly, and rebuild the statistic.
// Set its parent correctly.
old->Child(i).Parent() = this;
old->Child(i).Stat() = StatisticType(old->Child(i));
}
// Remove all the children so they don't get erased.
@@ -288,14 +306,19 @@ CoverTree<MetricType, StatisticType, MatType, RootPointPolicy>::CoverTree(
}
// Use the furthest descendant distance to determine the scale of the root
// node.
if (furthestDescendantDistance == 0.0)
// node. Note that if the root is a leaf, we can have scale INT_MIN, but if
// it *isn't* a leaf, we need to mark the scale as one higher than INT_MIN, so
// that the recursions don't fail.
if (furthestDescendantDistance == 0.0 && dataset->n_cols == 1)
scale = INT_MIN;
else if (furthestDescendantDistance == 0.0)
scale = INT_MIN + 1;
else
scale = (int) ceil(log(furthestDescendantDistance) / log(base));
// Initialize statistic.
stat = StatisticType(*this);
// Initialize statistics recursively after the entire tree construction is
// complete.
BuildStatistics<CoverTree, StatisticType>(this);
Log::Info << distanceComps << " distance computations during tree "
<< "construction." << std::endl;
@@ -363,9 +386,8 @@ CoverTree<MetricType, StatisticType, MatType, RootPointPolicy>::CoverTree(
{
children.push_back(&(old->Child(i)));
// Set its parent correctly, and rebuild the statistic.
// Set its parent correctly.
old->Child(i).Parent() = this;
old->Child(i).Stat() = StatisticType(old->Child(i));
}
// Remove all the children so they don't get erased.
@@ -379,14 +401,19 @@ CoverTree<MetricType, StatisticType, MatType, RootPointPolicy>::CoverTree(
}
// Use the furthest descendant distance to determine the scale of the root
// node.
if (furthestDescendantDistance == 0.0)
// node. Note that if the root is a leaf, we can have scale INT_MIN, but if
// it *isn't* a leaf, we need to mark the scale as one higher than INT_MIN, so
// that the recursions don't fail.
if (furthestDescendantDistance == 0.0 && dataset->n_cols == 1)
scale = INT_MIN;
else if (furthestDescendantDistance == 0.0)
scale = INT_MIN + 1;
else
scale = (int) ceil(log(furthestDescendantDistance) / log(base));
// Initialize statistic.
stat = StatisticType(*this);
// Initialize statistics recursively after the entire tree construction is
// complete.
BuildStatistics<CoverTree, StatisticType>(this);
Log::Info << distanceComps << " distance computations during tree "
<< "construction." << std::endl;
@@ -429,15 +456,11 @@ CoverTree<MetricType, StatisticType, MatType, RootPointPolicy>::CoverTree(
{
this->scale = INT_MIN;
numDescendants = 1;
stat = StatisticType(*this);
return;
}
// Otherwise, create the children.
CreateChildren(indices, distances, nearSetSize, farSetSize, usedSetSize);
// Initialize statistic.
stat = StatisticType(*this);
}
// Manually create a cover tree node.
@@ -472,9 +495,6 @@ CoverTree<MetricType, StatisticType, MatType, RootPointPolicy>::CoverTree(
// If necessary, create a local metric.
if (localMetric)
this->metric = new MetricType();
// Initialize the statistic.
stat = StatisticType(*this);
}
template<
@@ -907,8 +927,8 @@ CoverTree<MetricType, StatisticType, MatType, RootPointPolicy>::
other.Dataset().col(other.Point()));
math::RangeType<ElemType> result;
result.Lo() = distance - furthestDescendantDistance -
other.FurthestDescendantDistance();
result.Lo() = std::max(distance - furthestDescendantDistance -
other.FurthestDescendantDistance(), 0.0);
result.Hi() = distance + furthestDescendantDistance +
other.FurthestDescendantDistance();
@@ -930,8 +950,8 @@ CoverTree<MetricType, StatisticType, MatType, RootPointPolicy>::
const ElemType distance) const
{
math::RangeType<ElemType> result;
result.Lo() = distance - furthestDescendantDistance -
other.FurthestDescendantDistance();
result.Lo() = std::max(distance - furthestDescendantDistance -
other.FurthestDescendantDistance(), 0.0);
result.Hi() = distance + furthestDescendantDistance +
other.FurthestDescendantDistance();
@@ -952,8 +972,9 @@ CoverTree<MetricType, StatisticType, MatType, RootPointPolicy>::
{
const ElemType distance = metric->Evaluate(dataset->col(point), other);
return math::RangeType<ElemType>(distance - furthestDescendantDistance,
distance + furthestDescendantDistance);
return math::RangeType<ElemType>(
std::max(distance - furthestDescendantDistance, 0.0),
distance + furthestDescendantDistance);
}
//! Return the minimum and maximum distance to another point given that the
@@ -970,8 +991,9 @@ CoverTree<MetricType, StatisticType, MatType, RootPointPolicy>::
RangeDistance(const arma::vec& /* other */,
const ElemType distance) const
{
return math::RangeType<ElemType>(distance - furthestDescendantDistance,
distance + furthestDescendantDistance);
return math::RangeType<ElemType>(
std::max(distance - furthestDescendantDistance, 0.0),
distance + furthestDescendantDistance);
}
//! For a newly initialized node, create children using the near and far set.
@@ -1526,11 +1548,10 @@ inline void CoverTree<MetricType, StatisticType, MatType, RootPointPolicy>::
// Now take its child.
children.push_back(&(old->Child(0)));
// Set its parent and parameters correctly, and rebuild the statistic.
// Set its parent and parameters correctly.
old->Child(0).Parent() = this;
old->Child(0).ParentDistance() = old->ParentDistance();
old->Child(0).DistanceComps() = old->DistanceComps();
old->Child(0).Stat() = StatisticType(old->Child(0));
// Remove its child (so it doesn't delete it).
old->Children().erase(old->Children().begin() + old->Children().size() - 1);
src/mlpack/core/tree/octree/dual_tree_traverser_impl.hpp
+12
View File
@@ -42,6 +42,18 @@ void Octree<MetricType, StatisticType, MatType>::DualTreeTraverser<RuleType>::
// Store the current traversal info.
traversalInfo = rule.TraversalInfo();
// If both nodes are root nodes, just score them.
if (queryNode.Parent() == NULL && referenceNode.Parent() == NULL)
{
const double rootScore = rule.Score(queryNode, referenceNode);
// If root score is DBL_MAX, don't recurse.
if (rootScore == DBL_MAX)
{
++numPrunes;
return;
}
}
if (queryNode.IsLeaf() && referenceNode.IsLeaf())
{
const size_t begin = queryNode.Point(0);
src/mlpack/core/tree/octree/single_tree_traverser_impl.hpp
+14 -1
View File
@@ -22,7 +22,8 @@ template<typename MetricType, typename StatisticType, typename MatType>
template<typename RuleType>
Octree<MetricType, StatisticType, MatType>::SingleTreeTraverser<RuleType>::
SingleTreeTraverser(RuleType& rule) :
rule(rule)
rule(rule),
numPrunes(0)
{
// Nothing to do.
}
@@ -42,6 +43,18 @@ void Octree<MetricType, StatisticType, MatType>::SingleTreeTraverser<RuleType>::
}
else
{
// If it's the root node, just score it.
if (referenceNode.Parent() == NULL)
{
const double rootScore = rule.Score(queryIndex, referenceNode);
// If root score is DBL_MAX, don't recurse into that node.
if (rootScore == DBL_MAX)
{
++numPrunes;
return;
}
}
// Do a prioritized recursion, by scoring all candidates and then sorting
// them.
arma::vec scores(referenceNode.NumChildren());
src/mlpack/core/tree/rectangle_tree/rectangle_tree.hpp
+22 -1
View File
@@ -181,10 +181,24 @@ class RectangleTree
/**
* Create a rectangle tree by moving the other tree.
*
* @param other The tree to be copied.
* @param other The tree to be moved.
*/
RectangleTree(RectangleTree&& other);
/**
* Copy the given rectangle tree.
*
* @param other The tree to be copied.
*/
RectangleTree& operator=(const RectangleTree& other);
/**
* Take ownership of the given rectangle tree.
*
* @param other The tree to take ownership of.
*/
RectangleTree& operator=(RectangleTree&& other);
/**
* Construct the tree from a boost::serialization archive.
*/
@@ -542,6 +556,13 @@ class RectangleTree
*/
void SplitNode(std::vector<bool>& relevels);
/**
* Builds statistics for a node and all its descendants in a bottom-up way.
*
* @param node Node for which statistics will be built.
*/
void BuildStatistics(RectangleTree* node);
protected:
/**
* A default constructor. This is meant to only be used with
src/mlpack/core/tree/rectangle_tree/rectangle_tree_impl.hpp
+117 -5
View File
@@ -21,6 +21,25 @@
namespace mlpack {
namespace tree {
// Build the statistics, bottom-up.
template<typename MetricType,
typename StatisticType,
typename MatType,
typename SplitType,
typename DescentType,
template<typename> class AuxiliaryInformationType>
void RectangleTree<MetricType, StatisticType, MatType, SplitType, DescentType,
AuxiliaryInformationType>::
BuildStatistics(RectangleTree* node)
{
// Recurse first.
for (size_t i = 0; i < node->NumChildren(); ++i)
BuildStatistics(&node->Child(i));
// Now build the statistic.
node->Stat() = StatisticType(*node);
}
template<typename MetricType,
typename StatisticType,
typename MatType,
@@ -52,13 +71,14 @@ RectangleTree(const MatType& data,
points(maxLeafSize + 1), // Add one to make splitting the node simpler.
auxiliaryInfo(this)
{
stat = StatisticType(*this);
// For now, just insert the points in order.
RectangleTree* root = this;
for (size_t i = firstDataIndex; i < data.n_cols; i++)
root->InsertPoint(i);
// Initialize statistic recursively after tree construction is complete.
BuildStatistics(this);
}
template<typename MetricType,
@@ -92,13 +112,14 @@ RectangleTree(MatType&& data,
points(maxLeafSize + 1), // Add one to make splitting the node simpler.
auxiliaryInfo(this)
{
stat = StatisticType(*this);
// For now, just insert the points in order.
RectangleTree* root = this;
for (size_t i = firstDataIndex; i < dataset->n_cols; i++)
root->InsertPoint(i);
// Initialize statistic recursively after tree construction is complete.
BuildStatistics(this);
}
template<typename MetricType,
@@ -131,7 +152,8 @@ RectangleTree(
points(maxLeafSize + 1), // Add one to make splitting the node simpler.
auxiliaryInfo(this)
{
stat = StatisticType(*this);
// Initialize statistic.
BuildStatistics(this);
}
/**
@@ -234,6 +256,96 @@ RectangleTree(RectangleTree&& other) :
other.ownsDataset = false;
}
template<typename MetricType,
typename StatisticType,
typename MatType,
typename SplitType,
typename DescentType,
template<typename> class AuxiliaryInformationType>
RectangleTree<MetricType, StatisticType, MatType, SplitType, DescentType,
AuxiliaryInformationType>&
RectangleTree<MetricType, StatisticType, MatType, SplitType, DescentType,
AuxiliaryInformationType>::
operator=(const RectangleTree& other)
{
// Return if it's the same tree.
if (this == &other)
return *this;
for (size_t i = 0; i < numChildren; i++)
delete children[i];
if (ownsDataset)
delete dataset;
maxNumChildren = other.MaxNumChildren();
minNumChildren = other.MinNumChildren();
numChildren = other.NumChildren();
children.resize(maxNumChildren + 1, NULL);
parent = NULL;
begin = other.Begin();
count = other.Count();
numDescendants = other.numDescendants;
maxLeafSize = other.MaxLeafSize();
minLeafSize = other.MinLeafSize();
bound = other.bound;
parentDistance = other.ParentDistance();
dataset = new MatType(*other.dataset);
ownsDataset = true;
points = other.points;
auxiliaryInfo = AuxiliaryInfoType(other.auxiliaryInfo, this, true);
if (numChildren > 0)
{
for (size_t i = 0; i < numChildren; i++)
children[i] = new RectangleTree(other.Child(i), true, this);
}
return *this;
}
template<typename MetricType,
typename StatisticType,
typename MatType,
typename SplitType,
typename DescentType,
template<typename> class AuxiliaryInformationType>
RectangleTree<MetricType, StatisticType, MatType, SplitType, DescentType,
AuxiliaryInformationType>&
RectangleTree<MetricType, StatisticType, MatType, SplitType, DescentType,
AuxiliaryInformationType>::
operator=(RectangleTree&& other)
{
// Return if it's the same tree.
if (this == &other)
return *this;
for (size_t i = 0; i < numChildren; i++)
delete children[i];
if (ownsDataset)
delete dataset;
maxNumChildren = other.MaxNumChildren();
minNumChildren = other.MinNumChildren();
numChildren = other.NumChildren();
children = std::move(other.children);
parent = other.Parent();
begin = other.Begin();
count = other.Count();
numDescendants = other.numDescendants;
maxLeafSize = other.MaxLeafSize();
minLeafSize = other.MinLeafSize();
bound = std::move(other.bound);
parentDistance = other.ParentDistance();
dataset = other.dataset;
ownsDataset = other.ownsDataset;
points = std::move(other.points);
auxiliaryInfo = std::move(other.auxiliaryInfo);
return *this;
}
/**
* Construct the tree from a boost::serialization archive.
*/
src/mlpack/core/util/arma_config_check.hpp
+8 -14
View File
@@ -36,28 +36,22 @@ ARMA_64BIT_WORD disabled."
#endif
#endif
// Check if OpenMP was enabled when mlpack was built. This only matters for
// Armadillo 8.300.1 and newer.
#if (ARMA_VERSION_MAJOR > 8) || \
((ARMA_VERSION_MAJOR == 8) && (ARMA_VERSION_MINOR > 300)) || \
((ARMA_VERSION_MAJOR == 8) && (ARMA_VERSION_MINOR == 300) && \
(ARMA_VERSION_PATCH >= 1))
#ifdef ARMA_USE_OPENMP
#ifdef MLPACK_ARMA_DONT_USE_OPENMP
#pragma message "mlpack was compiled without OpenMP support, but you are \
// Check if OpenMP was enabled when mlpack was built.
#ifdef ARMA_USE_OPENMP
#ifdef MLPACK_ARMA_DONT_USE_OPENMP
#pragma message "mlpack was compiled without OpenMP support, but you are \
compiling with OpenMP support (either -fopenmp or another option). This will \
almost certainly cause irreparable disaster. Either compile your application \
*without* OpenMP support (i.e. remove -fopenmp or another flag), or, recompile \
mlpack with OpenMP support."
#endif
#else
#ifdef MLPACK_ARMA_USE_OPENMP
#pragma message "mlpack was compiled with OpenMP support, but you are \
#endif
#else
#ifdef MLPACK_ARMA_USE_OPENMP
#pragma message "mlpack was compiled with OpenMP support, but you are \
compiling without OpenMP support. This will almost certainly cause \
irreparable disaster. Either enable OpenMP support in your application (e.g., \
add -fopenmp to your compiler command line), or, recompile mlpack *without* \
OpenMP support."
#endif
#endif
#endif
src/mlpack/core/util/mlpack_main.hpp
+109
View File
@@ -36,11 +36,43 @@
#include <mlpack/bindings/cli/cli_option.hpp>
#include <mlpack/bindings/cli/print_doc_functions.hpp>
/**
* PRINT_PARAM_STRING() returns a string that contains the correct
* language-specific representation of a parameter's name.
*/
#define PRINT_PARAM_STRING mlpack::bindings::cli::ParamString
/**
* PRINT_PARAM_VALUE() returns a string that contains a correct
* language-specific representation of a parameter's value.
*/
#define PRINT_PARAM_VALUE mlpack::bindings::cli::PrintValue
/**
* PRINT_CALL() returns a string that contains the full language-specific
* representation of a call to an mlpack binding. The first argument should be
* the name of the binding, and all other arguments should be names of
* parameters followed by values (in the case where the preceding parameter is
* not a flag).
*/
#define PRINT_CALL mlpack::bindings::cli::ProgramCall
/**
* PRINT_DATASET() returns a string that contains a correct language-specific
* representation of a dataset name.
*/
#define PRINT_DATASET mlpack::bindings::cli::PrintDataset
/**
* PRINT_MODEL() returns a string that contains a correct language-specific
* representation of an mlpack model name.
*/
#define PRINT_MODEL mlpack::bindings::cli::PrintModel
/**
* BINDING_IGNORE_CHECK() is an internally-used macro to determine whether or
* not a specific parameter check should be ignored.
*/
#define BINDING_IGNORE_CHECK mlpack::bindings::cli::IgnoreCheck
namespace mlpack {
@@ -90,7 +122,20 @@ int main(int argc, char** argv)
#define PRINT_PARAM_VALUE(A, B) std::string(" ")
#define PRINT_DATASET(A) std::string(" ")
#define PRINT_MODEL(A) std::string(" ")
/**
* PRINT_CALL() returns a string that contains the full language-specific
* representation of a call to an mlpack binding. The first argument should be
* the name of the binding, and all other arguments should be names of
* parameters followed by values (in the case where the preceding parameter is
* not a flag).
*/
#define PRINT_CALL(...) std::string(" ")
/**
* BINDING_IGNORE_CHECK() is an internally-used macro to determine whether or
* not a specific parameter check should be ignored.
*/
#define BINDING_IGNORE_CHECK mlpack::bindings::tests::IgnoreCheck
namespace mlpack {
@@ -119,11 +164,43 @@ using Option = mlpack::bindings::tests::TestOption<T>;
#include <mlpack/bindings/python/py_option.hpp>
#include <mlpack/bindings/python/print_doc_functions.hpp>
/**
* PRINT_PARAM_STRING() returns a string that contains the correct
* language-specific representation of a parameter's name.
*/
#define PRINT_PARAM_STRING mlpack::bindings::python::ParamString
/**
* PRINT_PARAM_VALUE() returns a string that contains a correct
* language-specific representation of a parameter's value.
*/
#define PRINT_PARAM_VALUE mlpack::bindings::python::PrintValue
/**
* PRINT_DATASET() returns a string that contains a correct language-specific
* representation of a dataset name.
*/
#define PRINT_DATASET mlpack::bindings::python::PrintDataset
/**
* PRINT_MODEL() returns a string that contains a correct language-specific
* representation of an mlpack model name.
*/
#define PRINT_MODEL mlpack::bindings::python::PrintModel
/**
* PRINT_CALL() returns a string that contains the full language-specific
* representation of a call to an mlpack binding. The first argument should be
* the name of the binding, and all other arguments should be names of
* parameters followed by values (in the case where the preceding parameter is
* not a flag).
*/
#define PRINT_CALL mlpack::bindings::python::ProgramCall
/**
* BINDING_IGNORE_CHECK() is an internally-used macro to determine whether or
* not a specific parameter check should be ignored.
*/
#define BINDING_IGNORE_CHECK mlpack::bindings::python::IgnoreCheck
namespace mlpack {
@@ -170,11 +247,43 @@ PARAM_FLAG("copy_all_inputs", "If specified, all input parameters will be deep"
#include <mlpack/bindings/markdown/md_option.hpp>
#include <mlpack/bindings/markdown/print_doc_functions.hpp>
/**
* PRINT_PARAM_STRING() returns a string that contains the correct
* language-specific representation of a parameter's name.
*/
#define PRINT_PARAM_STRING mlpack::bindings::markdown::ParamString
/**
* PRINT_PARAM_VALUE() returns a string that contains a correct
* language-specific representation of a parameter's value.
*/
#define PRINT_PARAM_VALUE mlpack::bindings::markdown::PrintValue
/**
* PRINT_DATASET() returns a string that contains a correct language-specific
* representation of a dataset name.
*/
#define PRINT_DATASET mlpack::bindings::markdown::PrintDataset
/**
* PRINT_MODEL() returns a string that contains a correct language-specific
* representation of an mlpack model name.
*/
#define PRINT_MODEL mlpack::bindings::markdown::PrintModel
/**
* PRINT_CALL() returns a string that contains the full language-specific
* representation of a call to an mlpack binding. The first argument should be
* the name of the binding, and all other arguments should be names of
* parameters followed by values (in the case where the preceding parameter is
* not a flag).
*/
#define PRINT_CALL mlpack::bindings::markdown::ProgramCall
/**
* BINDING_IGNORE_CHECK() is an internally-used macro to determine whether or
* not a specific parameter check should be ignored.
*/
#define BINDING_IGNORE_CHECK mlpack::bindings::markdown::IgnoreCheck
// This doesn't actually matter for this binding type.
src/mlpack/core/util/param.hpp
+108 -36
View File
@@ -77,7 +77,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* Define a flag parameter.
*
* @param ID Name of the parameter.
* @param DESC Quick description of the parameter (1-2 sentences).
* @param DESC Quick description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @see mlpack::CLI, PROGRAM_INFO()
@@ -100,7 +102,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* --ID=value.
*
* @param ID Name of the parameter.
* @param DESC Quick description of the parameter (1-2 sentences).
* @param DESC Quick description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
* @param DEF Default value of the parameter.
*
@@ -131,7 +135,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* will be issued.
*
* @param ID Name of the parameter.
* @param DESC Quick description of the parameter (1-2 sentences).
* @param DESC Quick description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
*
* @see mlpack::CLI, PROGRAM_INFO()
*
@@ -153,7 +159,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* --ID=value.
*
* @param ID Name of the parameter.
* @param DESC Quick description of the parameter (1-2 sentences).
* @param DESC Quick description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
* @param DEF Default value of the parameter.
*
@@ -183,7 +191,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* will be issued.
*
* @param ID Name of the parameter.
* @param DESC Quick description of the parameter (1-2 sentences).
* @param DESC Quick description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
*
* @see mlpack::CLI, PROGRAM_INFO()
*
@@ -206,7 +216,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* PROGRAM_INFO() macro), the parameter can be specified with just --ID=value.
*
* @param ID Name of the parameter.
* @param DESC Quick description of the parameter (1-2 sentences).
* @param DESC Quick description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
* @param DEF Default value of the parameter.
*
@@ -236,7 +248,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* @endcode
*
* @param ID Name of the parameter.
* @param DESC Quick description of the parameter (1-2 sentences).
* @param DESC Quick description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @see mlpack::CLI, PROGRAM_INFO()
@@ -264,7 +278,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* @endcode
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -290,7 +306,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* @endcode
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -321,7 +339,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* types.
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -348,7 +368,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* @endcode
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -376,7 +398,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* @endcode
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -409,7 +433,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* types.
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -435,7 +461,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* @endcode
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -462,7 +490,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* @endcode
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -494,7 +524,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* types.
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -521,7 +553,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* @endcode
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -547,7 +581,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* @endcode
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -573,7 +609,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* @endcode
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -599,7 +637,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* @endcode
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -626,7 +666,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* @endcode
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -657,7 +699,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* types.
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -688,7 +732,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* types.
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -719,7 +765,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* types.
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -750,7 +798,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* types.
*
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @bug
@@ -771,7 +821,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* --ID=value1,value2,value3.
*
* @param ID Name of the parameter.
* @param DESC Quick description of the parameter (1-2 sentences).
* @param DESC Quick description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
* @param DEF Default value of the parameter.
*
@@ -802,7 +854,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* will be issued.
*
* @param ID Name of the parameter.
* @param DESC Quick description of the parameter (1-2 sentences).
* @param DESC Quick description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
*
* @see mlpack::CLI, PROGRAM_INFO()
*
@@ -838,7 +892,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* @endcode
*
* @param ID Name of the parameter.
* @param DESC Quick description of the parameter (1-2 sentences).
* @param DESC Quick description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS One-character string representing the alias of the parameter.
*
* @see mlpack::CLI, PROGRAM_INFO()
@@ -881,7 +937,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
*
* @param TYPE Type of the model to be loaded.
* @param ID Name of the parameter.
* @param DESC Description of the parameter.
* @param DESC Description of the parameter. Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*/
#define PARAM_MODEL_IN(TYPE, ID, DESC, ALIAS) \
@@ -913,7 +971,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
*
* @param TYPE Type of the model to be loaded.
* @param ID Name of the parameter.
* @param DESC Description of the parameter.
* @param DESC Description of the parameter. Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*/
#define PARAM_MODEL_IN_REQ(TYPE, ID, DESC, ALIAS) \
@@ -936,7 +996,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
*
* @param TYPE Type of the model to be saved.
* @param ID Name of the parameter.
* @param DESC Description of the parameter.
* @param DESC Description of the parameter. Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*/
#define PARAM_MODEL_OUT(TYPE, ID, DESC, ALIAS) \
@@ -948,7 +1010,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* The parameter must then be specified on the command line with --ID=value.
*
* @param ID Name of the parameter.
* @param DESC Quick description of the parameter (1-2 sentences).
* @param DESC Quick description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @see mlpack::CLI, PROGRAM_INFO()
@@ -970,7 +1034,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* The parameter must then be specified on the command line with --ID=value.
*
* @param ID Name of the parameter.
* @param DESC Quick description of the parameter (1-2 sentences).
* @param DESC Quick description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @see mlpack::CLI, PROGRAM_INFO()
@@ -992,7 +1058,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* The parameter must then be specified on the command line with --ID=value.
*
* @param ID Name of the parameter.
* @param DESC Quick description of the parameter (1-2 sentences).
* @param DESC Quick description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @see mlpack::CLI, PROGRAM_INFO()
@@ -1015,7 +1083,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
* --ID=value1,value2,value3.
*
* @param ID Name of the parameter.
* @param DESC Quick description of the parameter (1-2 sentences).
* @param DESC Quick description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS An alias for the parameter (one letter).
*
* @see mlpack::CLI, PROGRAM_INFO()
@@ -1052,7 +1122,9 @@ using DatasetInfo = DatasetMapper<IncrementPolicy, std::string>;
*
* @param T Type of the parameter.
* @param ID Name of the parameter.
* @param DESC Description of the parameter (1-2 sentences).
* @param DESC Description of the parameter (1-2 sentences). Don't use
* printing macros like PRINT_PARAM_STRING() or PRINT_DATASET() or others
* here---it will cause problems.
* @param ALIAS Alias for this parameter (one letter).
* @param DEF Default value of the parameter.
* @param REQ Whether or not parameter is required (boolean value).
src/mlpack/core/util/sfinae_utility.hpp
+4 -4
View File
@@ -220,7 +220,7 @@ struct NAME \
#define SINGLE_ARG(...) __VA_ARGS__
/**
* HAS_METHOD_FORM generates a template that allows to check at compile time
* HAS_METHOD_FORM generates a template that allows a compile time check for
* whether a given class has a method of the requested form. For example, for
* the following class
*
@@ -254,8 +254,8 @@ struct NAME \
HAS_METHOD_FORM_BASE(SINGLE_ARG(METHOD), SINGLE_ARG(NAME), 7)
/**
* HAS_EXACT_METHOD_FORM generates a template that allows to check at compile
* time whether a given class has a method of the requested form. For example,
* HAS_EXACT_METHOD_FORM generates a template that allows a compile time check
* whether a given class has a method of the requested form. For example,
* for the following class
*
* class A
@@ -291,7 +291,7 @@ struct NAME \
* A version of HAS_METHOD_FORM() where the maximum number of extra arguments is
* set to the default of 7.
*
* HAS_METHOD_FORM generates a template that allows to check at compile time
* HAS_METHOD_FORM generates a template that allows a compile time check as to
* whether a given class has a method of the requested form. For example, for
* the following class
*
src/mlpack/core/util/version.hpp
+1 -1
View File
@@ -17,7 +17,7 @@
// The version of mlpack. If this is a git repository, this will be a version
// with higher number than the most recent release.
#define MLPACK_VERSION_MAJOR 3
#define MLPACK_VERSION_MINOR 1
#define MLPACK_VERSION_MINOR 2
#define MLPACK_VERSION_PATCH 2
// The name of the version (for use by --version).
src/mlpack/methods/ann/CMakeLists.txt
+1
View File
@@ -18,6 +18,7 @@ add_subdirectory(convolution_rules)
add_subdirectory(gan)
add_subdirectory(rbm)
add_subdirectory(augmented)
add_subdirectory(regularizer)
# Add directory name to sources.
set(DIR_SRCS)
src/mlpack/methods/ann/ffn.hpp
+8
View File
@@ -257,6 +257,14 @@ class FFN
*/
void Add(LayerTypes<CustomLayers...> layer) { network.push_back(layer); }
//! Get the network model.
const std::vector<LayerTypes<CustomLayers...> >& Model() const
{
return network;
}
//! Modify the network model.
std::vector<LayerTypes<CustomLayers...> >& Model() { return network; }
//! Return the number of separable functions (the number of predictor points).
size_t NumFunctions() const { return numFunctions; }
src/mlpack/methods/ann/gan/CMakeLists.txt
+2
View File
@@ -8,6 +8,8 @@ set(SOURCES
wgangp_impl.hpp
)
add_subdirectory(metrics)
# Add directory name to sources.
set(DIR_SRCS)
foreach(file ${SOURCES})
src/mlpack/methods/ann/gan/gan.hpp
+1
View File
@@ -19,6 +19,7 @@
#include <mlpack/methods/ann/visitor/reset_visitor.hpp>
#include <mlpack/methods/ann/visitor/weight_size_visitor.hpp>
#include <mlpack/methods/ann/visitor/weight_set_visitor.hpp>
#include "metrics/inception_score.hpp"
namespace mlpack {
src/mlpack/methods/ann/gan/metrics/CMakeLists.txt
+15
View File
@@ -0,0 +1,15 @@
# Define the files we need to compile.
# Anything not in this list will not be compiled into mlpack.
set(SOURCES
inception_score
inception_score_impl
)
# Add directory name to sources.
set(DIR_SRCS)
foreach(file ${SOURCES})
set(DIR_SRCS ${DIR_SRCS} ${CMAKE_CURRENT_SOURCE_DIR}/${file})
endforeach()
# Append sources (with directory name) to list of all mlpack sources (used at
# the parent scope).
set(MLPACK_SRCS ${MLPACK_SRCS} ${DIR_SRCS} PARENT_SCOPE)
src/mlpack/methods/ann/gan/metrics/inception_score.hpp
+49
View File
@@ -0,0 +1,49 @@
/**
* @file inception_score.hpp
* @author Saksham Bansal
*
* Definition of Inception Score for Generative Adversarial Networks.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_METRICS_INCEPTION_SCORE_HPP
#define MLPACK_METHODS_METRICS_INCEPTION_SCORE_HPP
#include <mlpack/prereqs.hpp>
namespace mlpack {
namespace ann /* Artificial Neural Network */ {
/**
* Function that computes Inception Score for a set of images produced by a GAN.
* For more information, see the following.
*
* @code
* @article{Goodfellow2016,
* author = {Tim Salimans, Ian Goodfellow, Wojciech Zaremba, Vicki Cheung,
* Alec Radford, Xi Chen},
* title = {Improved Techniques for Training GANs},
* year = {2016},
* url = {https://arxiv.org/abs/1606.03498},
* }
* @endcode
*
* @param Model Model for evaluating the quality of images.
* @param images Images generated by GAN.
*/
template<typename ModelType>
double InceptionScore(ModelType Model,
arma::mat images,
size_t splits = 1);
} // namespace ann
} // namespace mlpack
#include "inception_score_impl.hpp"
#endif
src/mlpack/methods/ann/gan/metrics/inception_score_impl.hpp
+59
View File
@@ -0,0 +1,59 @@
/**
* @file inception_score_impl.hpp
* @author Saksham Bansal
*
* Definition of Inception Score for Generative Adversarial Networks.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_METRICS_INCEPTION_SCORE_IMPL_HPP
#define MLPACK_METHODS_METRICS_INCEPTION_SCORE_IMPL_HPP
// In case it hasn't been included yet.
#include "inception_score.hpp"
namespace mlpack {
namespace ann /* Artificial Neural Network */ {
template <typename ModelType>
double InceptionScore(ModelType model,
arma::mat images,
size_t splits)
{
size_t samples = images.n_cols;
size_t splitSize = samples / splits;
size_t remainder = samples % splits;
arma::mat preds;
model.Predict(images, preds);
size_t index = 0;
arma::vec scores = arma::vec(splits);
for (int i = 0; i < splits; i++)
{
size_t curSize = splitSize;
if (remainder)
{
curSize++;
remainder--;
}
arma::mat curPreds =
arma::mat(preds.colptr(index), preds.n_rows, curSize, false, true);
arma::colvec c = arma::log(arma::mean(curPreds, 1));
arma::mat temp = arma::log(curPreds);
temp.each_col() -= c;
curPreds %= temp;
scores(i) = exp(arma::as_scalar(arma::mean(arma::sum(curPreds, 0))));
index += curSize;
}
return arma::mean(scores);
}
} // namespace ann
} // namespace mlpack
#endif
src/mlpack/methods/ann/layer/CMakeLists.txt
+6
View File
@@ -40,6 +40,8 @@ set(SOURCES
gru_impl.hpp
hard_tanh.hpp
hard_tanh_impl.hpp
highway.hpp
highway_impl.hpp
join.hpp
join_impl.hpp
layer.hpp
@@ -63,6 +65,8 @@ set(SOURCES
max_pooling_impl.hpp
mean_pooling.hpp
mean_pooling_impl.hpp
minibatch_discrimination.hpp
minibatch_discrimination_impl.hpp
multiply_constant.hpp
multiply_constant_impl.hpp
multiply_merge.hpp
@@ -88,6 +92,8 @@ set(SOURCES
vr_class_reward_impl.hpp
c_relu.hpp
c_relu_impl.hpp
weight_norm.hpp
weight_norm_impl.hpp
)
# Add directory name to sources.
src/mlpack/methods/ann/layer/alpha_dropout.hpp
-1
View File
@@ -37,7 +37,6 @@ namespace ann /** Artificial Neural Network. */ {
* journal = {Advances in Neural Information Processing Systems},
* year = {2017}
* }
* }
* @endcode
*
* @tparam InputDataType Type of the input data (arma::colvec, arma::mat,
src/mlpack/methods/ann/layer/atrous_convolution.hpp
+7 -47
View File
@@ -21,6 +21,7 @@
#include <mlpack/methods/ann/convolution_rules/svd_convolution.hpp>
#include "layer_types.hpp"
#include "padding.hpp"
namespace mlpack{
namespace ann /** Artificial Neural Network. */ {
@@ -165,6 +166,9 @@ class AtrousConvolution
//! Modify the output height.
size_t& OutputHeight() { return outputHeight; }
//! Modify the bias weights of the layer.
arma::mat& Bias() { return bias; }
/**
* Serialize the layer
*/
@@ -220,53 +224,6 @@ class AtrousConvolution
output = arma::fliplr(arma::flipud(input));
}
/*
* Pad the given input data.
*
* @param input The input to be padded.
* @param wPad Padding width of the input.
* @param hPad Padding height of the input.
* @param output The padded output data.
*/
template<typename eT>
void Pad(const arma::Mat<eT>& input,
size_t wPad,
size_t hPad,
arma::Mat<eT>& output)
{
if (output.n_rows != input.n_rows + wPad * 2 ||
output.n_cols != input.n_cols + hPad * 2)
{
output = arma::zeros(input.n_rows + wPad * 2, input.n_cols + hPad * 2);
}
output.submat(wPad, hPad, wPad + input.n_rows - 1,
hPad + input.n_cols - 1) = input;
}
/*
* Pad the given input data.
*
* @param input The input to be padded.
* @param wPad Padding width of the input.
* @param hPad Padding height of the input.
* @param output The padded output data.
*/
template<typename eT>
void Pad(const arma::Cube<eT>& input,
size_t wPad,
size_t hPad,
arma::Cube<eT>& output)
{
output = arma::zeros(input.n_rows + wPad * 2,
input.n_cols + hPad * 2, input.n_slices);
for (size_t i = 0; i < input.n_slices; ++i)
{
Pad<double>(input.slice(i), wPad, hPad, output.slice(i));
}
}
//! Locally-stored number of input channels.
size_t inSize;
@@ -336,6 +293,9 @@ class AtrousConvolution
//! Locally-stored transformed gradient parameter.
arma::cube gradientTemp;
//! Locally-stored padding layer.
Padding<>* padding;
//! Locally-stored delta object.
OutputDataType delta;
src/mlpack/methods/ann/layer/atrous_convolution_impl.hpp
+9 -1
View File
@@ -79,6 +79,7 @@ AtrousConvolution<
dilationH(dilationH)
{
weights.set_size((outSize * inSize * kW * kH) + outSize, 1);
padding = new Padding<>(padW, padW, padH, padH);
}
template<
@@ -124,7 +125,14 @@ void AtrousConvolution<
if (padW != 0 || padH != 0)
{
Pad(inputTemp, padW, padH, inputPaddedTemp);
inputPaddedTemp.set_size(inputTemp.n_rows + padW * 2,
inputTemp.n_cols + padH * 2, inputTemp.n_slices);
for (size_t i = 0; i < inputTemp.n_slices; ++i)
{
padding->Forward(std::move(inputTemp.slice(i)),
std::move(inputPaddedTemp.slice(i)));
}
}
size_t wConv = ConvOutSize(inputWidth, kW, dW, padW, dilationW);
src/mlpack/methods/ann/layer/batch_norm.hpp
+4 -1
View File
@@ -151,7 +151,7 @@ class BatchNorm
//! Locally-stored epsilon value.
double eps;
//! Whether we are in loading or saving mode.
//! Variable to keep track of whether we are in loading or saving mode.
bool loading;
//! Locally-stored scale parameter.
@@ -195,6 +195,9 @@ class BatchNorm
//! Locally-stored normalized input.
OutputDataType normalized;
//! Locally-stored zero mean input.
OutputDataType inputMean;
}; // class BatchNorm
} // namespace ann
src/mlpack/methods/ann/layer/batch_norm_impl.hpp
+7 -7
View File
@@ -22,6 +22,7 @@ namespace ann { /** Artificial Neural Network. */
template<typename InputDataType, typename OutputDataType>
BatchNorm<InputDataType, OutputDataType>::BatchNorm() :
size(0),
eps(1e-8),
loading(false),
deterministic(false),
@@ -80,6 +81,7 @@ void BatchNorm<InputDataType, OutputDataType>::Forward(
// Normalize the input.
output = input.each_col() - mean;
inputMean = output;
output.each_col() /= arma::sqrt(variance + eps);
// Use Welford method to compute the sample variance and mean.
@@ -87,9 +89,9 @@ void BatchNorm<InputDataType, OutputDataType>::Forward(
{
count += 1;
OutputDataType delta = input.col(i) - runningMean;
runningMean = runningMean + delta / count;
runningVariance += delta % (input.col(i) - runningMean);
OutputDataType diff = input.col(i) - runningMean;
runningMean = runningMean + diff / count;
runningVariance += diff % (input.col(i) - runningMean);
}
// Reused in the backward and gradient step.
@@ -106,10 +108,9 @@ template<typename eT>
void BatchNorm<InputDataType, OutputDataType>::Backward(
const arma::Mat<eT>&& input, arma::Mat<eT>&& gy, arma::Mat<eT>&& g)
{
const arma::mat inputMean = input.each_col() - mean;
const arma::mat stdInv = 1.0 / arma::sqrt(variance + eps);
// Step 1: dl / dxhat
// Step 1: dl / dxhat.
const arma::mat norm = gy.each_col() % gamma;
// Step 2: sum dl / dxhat * (x - mu) * -0.5 * stdInv^3.
@@ -123,8 +124,7 @@ void BatchNorm<InputDataType, OutputDataType>::Backward(
// Step 3: sum (dl / dxhat * -1 / stdInv) + variance *
// (sum -2 * (x - mu)) / m.
g.each_col() += (arma::sum(norm.each_col() % -stdInv, 1) + (var %
arma::mean(-2 * inputMean, 1))) / input.n_cols;
g.each_col() += arma::sum(norm.each_col() % -stdInv, 1) / input.n_cols;
}
template<typename InputDataType, typename OutputDataType>
src/mlpack/methods/ann/layer/convolution.hpp
+7 -47
View File
@@ -20,6 +20,7 @@
#include <mlpack/methods/ann/convolution_rules/svd_convolution.hpp>
#include "layer_types.hpp"
#include "padding.hpp"
namespace mlpack {
namespace ann /** Artificial Neural Network. */ {
@@ -161,6 +162,9 @@ class Convolution
//! Modify the output height.
size_t& OutputHeight() { return outputHeight; }
//! Modify the bias weights of the layer.
arma::mat& Bias() { return bias; }
/**
* Serialize the layer
*/
@@ -214,53 +218,6 @@ class Convolution
output = arma::fliplr(arma::flipud(input));
}
/*
* Pad the given input data.
*
* @param input The input to be padded.
* @param wPad Padding width of the input.
* @param hPad Padding height of the input.
* @param output The padded output data.
*/
template<typename eT>
void Pad(const arma::Mat<eT>& input,
size_t wPad,
size_t hPad,
arma::Mat<eT>& output)
{
if (output.n_rows != input.n_rows + wPad * 2 ||
output.n_cols != input.n_cols + hPad * 2)
{
output = arma::zeros(input.n_rows + wPad * 2, input.n_cols + hPad * 2);
}
output.submat(wPad, hPad, wPad + input.n_rows - 1,
hPad + input.n_cols - 1) = input;
}
/*
* Pad the given input data.
*
* @param input The input to be padded.
* @param wPad Padding width of the input.
* @param hPad Padding height of the input.
* @param output The padded output data.
*/
template<typename eT>
void Pad(const arma::Cube<eT>& input,
size_t wPad,
size_t hPad,
arma::Cube<eT>& output)
{
output = arma::zeros(input.n_rows + wPad * 2,
input.n_cols + hPad * 2, input.n_slices);
for (size_t i = 0; i < input.n_slices; ++i)
{
Pad<eT>(input.slice(i), wPad, hPad, output.slice(i));
}
}
//! Locally-stored number of input channels.
size_t inSize;
@@ -324,6 +281,9 @@ class Convolution
//! Locally-stored transformed gradient parameter.
arma::cube gradientTemp;
//! Locally-stored padding layer.
Padding<>* padding;
//! Locally-stored delta object.
OutputDataType delta;
src/mlpack/methods/ann/layer/convolution_impl.hpp
+9 -1
View File
@@ -74,6 +74,7 @@ Convolution<
outputHeight(0)
{
weights.set_size((outSize * inSize * kW * kH) + outSize, 1);
padding = new Padding<>(padW, padW, padH, padH);
}
template<
@@ -119,7 +120,14 @@ void Convolution<
if (padW != 0 || padH != 0)
{
Pad(inputTemp, padW, padH, inputPaddedTemp);
inputPaddedTemp.set_size(inputTemp.n_rows + padW * 2,
inputTemp.n_cols + padH * 2, inputTemp.n_slices);
for (size_t i = 0; i < inputTemp.n_slices; ++i)
{
padding->Forward(std::move(inputTemp.slice(i)),
std::move(inputPaddedTemp.slice(i)));
}
}
size_t wConv = ConvOutSize(inputWidth, kW, dW, padW);
src/mlpack/methods/ann/layer/highway.hpp
+257
View File
@@ -0,0 +1,257 @@
/**
* @file highway.hpp
* @author Konstantin Sidorov
* @author Saksham Bansal
*
* Definition of the Highway layer.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_ANN_LAYER_HIGHWAY_HPP
#define MLPACK_METHODS_ANN_LAYER_HIGHWAY_HPP
#include <mlpack/prereqs.hpp>
#include <boost/ptr_container/ptr_vector.hpp>
#include "../visitor/delete_visitor.hpp"
#include "../visitor/delta_visitor.hpp"
#include "../visitor/output_height_visitor.hpp"
#include "../visitor/output_parameter_visitor.hpp"
#include "../visitor/output_width_visitor.hpp"
#include "layer_types.hpp"
#include "add_merge.hpp"
namespace mlpack {
namespace ann /** Artificial Neural Network. */ {
/**
* Implementation of the Highway layer. The Highway class can vary its behavior
* between that of feed-forward fully connected network container and that
* of a layer which simply passes its inputs through depending on the transform
* gate. Note that the size of the input and output matrices of this class
* should be equal.
*
* For more information, refer the following paper.
*
* @code
* @article{Srivastava2015,
* author = {Rupesh Kumar Srivastava, Klaus Greff, Jurgen Schmidhuber},
* title = {Training Very Deep Networks},
* journal = {Advances in Neural Information Processing Systems},
* year = {2015},
* url = {https://arxiv.org/abs/1507.06228},
* }
* @endcode
*
* @tparam InputDataType Type of the input data (arma::colvec, arma::mat,
* arma::sp_mat or arma::cube).
* @tparam OutputDataType Type of the output data (arma::colvec, arma::mat,
* arma::sp_mat or arma::cube).
*/
template <
typename InputDataType = arma::mat,
typename OutputDataType = arma::mat,
typename... CustomLayers>
class Highway
{
public:
//! Create the Highway object.
Highway();
/**
* Create the Highway object.
*
* @param inSize The number of input units.
* @param model Expose all the network modules.
*/
Highway(const size_t inSize, const bool model = true);
//! Destroy the Highway object.
~Highway();
/**
* Destroy all the modules added to the Highway object.
*/
void DeleteModules();
/**
* Reset the layer parameter.
*/
void Reset();
/**
* Ordinary feed-forward pass of a neural network, evaluating the function
* f(x) by propagating the activity forward through f.
*
* @param input Input data used for evaluating the specified function.
* @param output Resulting output activation.
*/
template<typename eT>
void Forward(arma::Mat<eT>&& input, arma::Mat<eT>&& output);
/**
* Ordinary feed-backward pass of a neural network, calculating the function
* f(x) by propagating x backwards through f. Using the results from the
* feed-forward pass.
*
* @param input The propagated input activation.
* @param gy The backpropagated error.
* @param g The calculated gradient.
*/
template<typename eT>
void Backward(const arma::Mat<eT>&& /* input */,
arma::Mat<eT>&& gy,
arma::Mat<eT>&& g);
/**
* Calculate the gradient using the output delta and the input activation.
*
* @param input The input parameter used for calculating the gradient.
* @param error The calculated error.
* @param gradient The calculated gradient.
*/
template<typename eT>
void Gradient(arma::Mat<eT>&& input,
arma::Mat<eT>&& error,
arma::Mat<eT>&& gradient);
/**
* Add a new module to the model.
*
* @param args The layer parameter.
*/
template <class LayerType, class... Args>
void Add(Args... args) { network.push_back(new LayerType(args...)); }
/**
* Add a new module to the model.
*
* @param layer The Layer to be added to the model.
*/
void Add(LayerTypes<CustomLayers...> layer) { network.push_back(layer); }
//! Return the modules of the model.
std::vector<LayerTypes<CustomLayers...> >& Model()
{
if (model)
{
return network;
}
return empty;
}
//! Get the parameters.
OutputDataType const& Parameters() const { return weights; }
//! Modify the parameters.
OutputDataType& Parameters() { return weights; }
//! Get the input parameter.
InputDataType const& InputParameter() const { return inputParameter; }
//! Modify the input parameter.
InputDataType& InputParameter() { return inputParameter; }
//! Get the output parameter.
OutputDataType const& OutputParameter() const { return outputParameter; }
//! Modify the output parameter.
OutputDataType& OutputParameter() { return outputParameter; }
//! Get the delta.
OutputDataType const& Delta() const { return delta; }
//! Modify the delta.
OutputDataType& Delta() { return delta; }
//! Get the gradient.
OutputDataType const& Gradient() const { return gradient; }
//! Modify the gradient.
OutputDataType& Gradient() { return gradient; }
/**
* Serialize the layer.
*/
template<typename Archive>
void serialize(Archive& ar, const unsigned int /* version */);
private:
//! Locally-stored number of input units.
size_t inSize;
//! Parameter which indicates if the modules should be exposed.
bool model;
//! Indicator if we already initialized the model.
bool reset;
//! Locally-stored network modules.
std::vector<LayerTypes<CustomLayers...> > network;
//! Locally-stored empty list of modules.
std::vector<LayerTypes<CustomLayers...> > empty;
//! Locally-stored weight object.
OutputDataType weights;
//! Locally-stored delta object.
OutputDataType delta;
//! Locally-stored gradient object.
OutputDataType gradient;
//! Weights for transformation of output.
OutputDataType transformWeight;
//! Bias for transformation of output.
OutputDataType transformBias;
//! Locally-stored transform gate parameters.
OutputDataType transformGate;
//! Locally-stored transform gate activation.
OutputDataType transformGateActivation;
//! Locally-stored transform gate error.
OutputDataType transformGateError;
//! Locally-stored input parameter object.
InputDataType inputParameter;
//! Locally-stored output parameter object.
OutputDataType outputParameter;
//! The input width.
size_t width;
//! The input height.
size_t height;
//! The normal output without highway network.
OutputDataType networkOutput;
//! Locally-stored delta visitor.
DeltaVisitor deltaVisitor;
//! Locally-stored output parameter visitor.
OutputParameterVisitor outputParameterVisitor;
//! Locally-stored delete visitor.
DeleteVisitor deleteVisitor;
//! Locally-stored output width visitor.
OutputWidthVisitor outputWidthVisitor;
//! Locally-stored output height visitor.
OutputHeightVisitor outputHeightVisitor;
}; // class Highway
} // namespace ann
} // namespace mlpack
// Include implementation.
#include "highway_impl.hpp"
#endif
src/mlpack/methods/ann/layer/highway_impl.hpp
+249
View File
@@ -0,0 +1,249 @@
/**
* @file highway_impl.hpp
* @author Konstantin Sidorov
* @author Saksham Bansal
*
* Implementation of Highway layer.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_ANN_LAYER_HIGHWAY_IMPL_HPP
#define MLPACK_METHODS_ANN_LAYER_HIGHWAY_IMPL_HPP
// In case it hasn't yet been included.
#include "highway.hpp"
#include "../visitor/forward_visitor.hpp"
#include "../visitor/backward_visitor.hpp"
#include "../visitor/gradient_visitor.hpp"
#include "../visitor/set_input_height_visitor.hpp"
#include "../visitor/set_input_width_visitor.hpp"
namespace mlpack {
namespace ann /** Artificial Neural Network. */ {
template<typename InputDataType, typename OutputDataType,
typename... CustomLayers>
Highway<InputDataType, OutputDataType, CustomLayers...>::Highway() :
inSize(0),
model(true),
reset(false),
width(0),
height(0)
{
// Nothing to do here.
}
template<
typename InputDataType, typename OutputDataType, typename... CustomLayers>
Highway<InputDataType, OutputDataType, CustomLayers...>::Highway(
const size_t inSize,
const bool model) :
inSize(inSize),
model(model),
reset(false),
width(0),
height(0)
{
weights.set_size(inSize * inSize + inSize, 1);
}
template<typename InputDataType, typename OutputDataType,
typename... CustomLayers>
Highway<InputDataType, OutputDataType, CustomLayers...>::~Highway()
{
if (!model)
{
for (LayerTypes<CustomLayers...>& layer : network)
{
boost::apply_visitor(deleteVisitor, layer);
}
}
}
template<typename InputDataType, typename OutputDataType,
typename... CustomLayers>
void Highway<
InputDataType, OutputDataType, CustomLayers...>::DeleteModules()
{
if (model)
{
for (LayerTypes<CustomLayers...>& layer : network)
{
boost::apply_visitor(deleteVisitor, layer);
}
}
}
template<typename InputDataType, typename OutputDataType,
typename... CustomLayers>
void Highway<InputDataType, OutputDataType, CustomLayers...>::Reset()
{
transformWeight = arma::mat(weights.memptr(), inSize, inSize, false, false);
transformBias = arma::mat(weights.memptr() + transformWeight.n_elem,
inSize, 1, false, false);
}
template<typename InputDataType, typename OutputDataType,
typename... CustomLayers>
template<typename eT>
void Highway<InputDataType, OutputDataType, CustomLayers...>::Forward(
arma::Mat<eT>&& input, arma::Mat<eT>&& output)
{
boost::apply_visitor(ForwardVisitor(std::move(input), std::move(
boost::apply_visitor(outputParameterVisitor, network.front()))),
network.front());
if (!reset)
{
if (boost::apply_visitor(outputWidthVisitor, network.front()) != 0)
{
width = boost::apply_visitor(outputWidthVisitor, network.front());
}
if (boost::apply_visitor(outputHeightVisitor, network.front()) != 0)
{
height = boost::apply_visitor(outputHeightVisitor, network.front());
}
}
for (size_t i = 1; i < network.size(); ++i)
{
if (!reset)
{
// Set the input width.
boost::apply_visitor(SetInputWidthVisitor(width), network[i]);
// Set the input height.
boost::apply_visitor(SetInputHeightVisitor(height), network[i]);
}
boost::apply_visitor(ForwardVisitor(std::move(boost::apply_visitor(
outputParameterVisitor, network[i - 1])), std::move(
boost::apply_visitor(outputParameterVisitor, network[i]))),
network[i]);
if (!reset)
{
// Get the output width.
if (boost::apply_visitor(outputWidthVisitor, network[i]) != 0)
{
width = boost::apply_visitor(outputWidthVisitor, network[i]);
}
// Get the output height.
if (boost::apply_visitor(outputHeightVisitor, network[i]) != 0)
{
height = boost::apply_visitor(outputHeightVisitor, network[i]);
}
}
}
if (!reset)
{
reset = true;
}
output = boost::apply_visitor(outputParameterVisitor, network.back());
if (arma::size(output) != arma::size(input))
{
Log::Fatal << "The sizes of the output and input matrices of the Highway"
<< " network should be equal. Please examine the network layers.";
}
transformGate = transformWeight * input;
transformGate.each_col() += transformBias;
transformGateActivation = 1.0 /(1 + arma::exp(-transformGate));
inputParameter = input;
networkOutput = output;
output = (output % transformGateActivation) +
(input % (1 - transformGateActivation));
}
template<typename InputDataType, typename OutputDataType,
typename... CustomLayers>
template<typename eT>
void Highway<InputDataType, OutputDataType, CustomLayers...>::Backward(
const arma::Mat<eT>&& /* input */,
arma::Mat<eT>&& gy,
arma::Mat<eT>&& g)
{
boost::apply_visitor(BackwardVisitor(std::move(boost::apply_visitor(
outputParameterVisitor, network.back())),
std::move(gy % transformGateActivation),
std::move(boost::apply_visitor(deltaVisitor, network.back()))),
network.back());
for (size_t i = 2; i < network.size() + 1; ++i)
{
boost::apply_visitor(BackwardVisitor(std::move(boost::apply_visitor(
outputParameterVisitor, network[network.size() - i])), std::move(
boost::apply_visitor(deltaVisitor, network[network.size() - i + 1])),
std::move(boost::apply_visitor(deltaVisitor,
network[network.size() - i]))), network[network.size() - i]);
}
g = boost::apply_visitor(deltaVisitor, network.front());
transformGateError = gy % (networkOutput - inputParameter) %
transformGateActivation % (1.0 - transformGateActivation);
g += transformWeight.t() * transformGateError;
g += gy % (1 - transformGateActivation);
}
template<typename InputDataType, typename OutputDataType,
typename... CustomLayers>
template<typename eT>
void Highway<InputDataType, OutputDataType, CustomLayers...>::Gradient(
arma::Mat<eT>&& input,
arma::Mat<eT>&& error,
arma::Mat<eT>&& gradient)
{
boost::apply_visitor(GradientVisitor(std::move(boost::apply_visitor(
outputParameterVisitor, network[network.size() - 2])),
std::move(error % transformGateActivation)), network.back());
for (size_t i = 2; i < network.size(); ++i)
{
boost::apply_visitor(GradientVisitor(std::move(boost::apply_visitor(
outputParameterVisitor, network[network.size() - i - 1])), std::move(
boost::apply_visitor(deltaVisitor, network[network.size() - i + 1]))),
network[network.size() - i]);
}
boost::apply_visitor(GradientVisitor(std::move(input), std::move(
boost::apply_visitor(deltaVisitor, network[1]))), network.front());
gradient.submat(0, 0, transformWeight.n_elem - 1, 0) = arma::vectorise(
transformGateError * input.t());
gradient.submat(transformWeight.n_elem, 0, gradient.n_elem - 1, 0) =
arma::sum(transformGateError, 1);
}
template<typename InputDataType, typename OutputDataType,
typename... CustomLayers>
template<typename Archive>
void Highway<InputDataType, OutputDataType, CustomLayers...>::serialize(
Archive& ar, const unsigned int /* version */)
{
// If loading, delete the old layers and set size for weights.
if (Archive::is_loading::value)
{
for (LayerTypes<CustomLayers...>& layer : network)
{
boost::apply_visitor(deleteVisitor, layer);
}
weights.set_size(inSize * inSize + inSize, 1);
}
ar & BOOST_SERIALIZATION_NVP(model);
ar & BOOST_SERIALIZATION_NVP(network);
}
} // namespace ann
} // namespace mlpack
#endif
src/mlpack/methods/ann/layer/layer.hpp
+4
View File
@@ -19,12 +19,15 @@
#include "convolution.hpp"
#include "dropconnect.hpp"
#include "glimpse.hpp"
#include "highway.hpp"
#include "layer_norm.hpp"
#include "layer_types.hpp"
#include "linear.hpp"
#include "linear_no_bias.hpp"
#include "lstm.hpp"
#include "minibatch_discrimination.hpp"
#include "multiply_merge.hpp"
#include "padding.hpp"
#include "gru.hpp"
#include "fast_lstm.hpp"
#include "recurrent.hpp"
@@ -35,5 +38,6 @@
#include "concat.hpp"
#include "vr_class_reward.hpp"
#include "transposed_convolution.hpp"
#include "weight_norm.hpp"
#endif
src/mlpack/methods/ann/layer/layer_norm.hpp
+4 -1
View File
@@ -155,7 +155,7 @@ class LayerNorm
//! Locally-stored epsilon value.
double eps;
//! Whether we are in loading or saving mode.
//! Variable to keep track of whether we are in loading or saving mode.
bool loading;
//! Locally-stored scale parameter.
@@ -184,6 +184,9 @@ class LayerNorm
//! Locally-stored normalized input.
OutputDataType normalized;
//! Locally-stored zero mean input.
OutputDataType inputMean;
}; // class LayerNorm
} // namespace ann
src/mlpack/methods/ann/layer/layer_norm_impl.hpp
+4 -5
View File
@@ -22,6 +22,7 @@ namespace ann { /** Artificial Neural Network. */
template<typename InputDataType, typename OutputDataType>
LayerNorm<InputDataType, OutputDataType>::LayerNorm() :
size(0),
eps(1e-8),
loading(false)
{
@@ -63,7 +64,7 @@ void LayerNorm<InputDataType, OutputDataType>::Forward(
// Normalize the input.
output = input.each_row() - mean;
inputMean = output;
output.each_row() /= arma::sqrt(variance + eps);
// Reused in the backward and gradient step.
@@ -79,10 +80,9 @@ template<typename eT>
void LayerNorm<InputDataType, OutputDataType>::Backward(
const arma::Mat<eT>&& input, arma::Mat<eT>&& gy, arma::Mat<eT>&& g)
{
const arma::mat inputMean = input.each_row() - mean;
const arma::mat stdInv = 1.0 / arma::sqrt(variance + eps);
// dl / dxhat
// dl / dxhat.
const arma::mat norm = gy.each_col() % gamma;
// sum dl / dxhat * (x - mu) * -0.5 * stdInv^3.
@@ -96,8 +96,7 @@ void LayerNorm<InputDataType, OutputDataType>::Backward(
// sum (dl / dxhat * -1 / stdInv) + variance *
// (sum -2 * (x - mu)) / m.
g.each_row() += (arma::sum(norm.each_row() % -stdInv, 0) + (var %
arma::mean(-2 * inputMean, 0))) / input.n_rows;
g.each_row() += arma::sum(norm.each_row() % -stdInv, 0) / input.n_rows;
}
template<typename InputDataType, typename OutputDataType>
src/mlpack/methods/ann/layer/layer_traits.hpp
+4
View File
@@ -112,6 +112,10 @@ HAS_MEM_FUNC(Loss, HasLoss);
// can use with SFINAE to catch when a type has a Run() function.
HAS_MEM_FUNC(Run, HasRunCheck);
// This gives us a HasBiasCheck<T, U> type (where U is a function pointer) we
// can use with SFINAE to catch when a type has a Bias() function.
HAS_MEM_FUNC(Bias, HasBiasCheck);
} // namespace ann
} // namespace mlpack
src/mlpack/methods/ann/layer/layer_types.hpp
+57 -13
View File
@@ -40,12 +40,16 @@
#include <mlpack/methods/ann/layer/reparametrization.hpp>
#include <mlpack/methods/ann/layer/select.hpp>
#include <mlpack/methods/ann/layer/subview.hpp>
#include <mlpack/methods/ann/layer/virtual_batch_norm.hpp>
// Convolution modules.
#include <mlpack/methods/ann/convolution_rules/border_modes.hpp>
#include <mlpack/methods/ann/convolution_rules/naive_convolution.hpp>
#include <mlpack/methods/ann/convolution_rules/fft_convolution.hpp>
// Regularizers.
#include <mlpack/methods/ann/regularizer/no_regularizer.hpp>
// Loss function modules.
#include <mlpack/methods/ann/loss_functions/negative_log_likelihood.hpp>
@@ -56,13 +60,32 @@ template<typename InputDataType, typename OutputDataType> class BatchNorm;
template<typename InputDataType, typename OutputDataType> class DropConnect;
template<typename InputDataType, typename OutputDataType> class Glimpse;
template<typename InputDataType, typename OutputDataType> class LayerNorm;
template<typename InputDataType, typename OutputDataType> class Linear;
template<typename InputDataType, typename OutputDataType> class LinearNoBias;
template<typename InputDataType, typename OutputDataType> class LSTM;
template<typename InputDataType, typename OutputDataType> class GRU;
template<typename InputDataType, typename OutputDataType> class FastLSTM;
template<typename InputDataType, typename OutputDataType> class VRClassReward;
template<typename InputDataType, typename OutputDataType> class Concatenate;
template<typename InputDataType, typename OutputDataType> class Padding;
template<typename InputDataType,
typename OutputDataType,
typename RegularizerType>
class Linear;
template<typename InputDataType,
typename OutputDataType,
typename RegularizerType>
class LinearNoBias;
template<typename InputDataType,
typename OutputDataType
>
class VirtualBatchNorm;
template<typename InputDataType,
typename OutputDataType
>
class MiniBatchDiscrimination;
template<typename InputDataType,
typename OutputDataType
@@ -82,6 +105,12 @@ template<typename InputDataType,
>
class Sequential;
template<typename InputDataType,
typename OutputDataType,
typename... CustomLayers
>
class Highway;
template<typename InputDataType,
typename OutputDataType,
typename... CustomLayers
@@ -140,6 +169,26 @@ template<typename InputDataType,
>
class MultiplyMerge;
template <typename InputDataType,
typename OutputDataType,
typename... CustomLayers
>
class WeightNorm;
using MoreTypes = boost::variant<
Recurrent<arma::mat, arma::mat>*,
RecurrentAttention<arma::mat, arma::mat>*,
ReinforceNormal<arma::mat, arma::mat>*,
Reparametrization<arma::mat, arma::mat>*,
Select<arma::mat, arma::mat>*,
Sequential<arma::mat, arma::mat, false>*,
Sequential<arma::mat, arma::mat, true>*,
Subview<arma::mat, arma::mat>*,
VRClassReward<arma::mat, arma::mat>*,
VirtualBatchNorm<arma::mat, arma::mat>*,
WeightNorm<arma::mat, arma::mat>*
>;
template <typename... CustomLayers>
using LayerTypes = boost::variant<
Add<arma::mat, arma::mat>*,
@@ -173,12 +222,13 @@ using LayerTypes = boost::variant<
FlexibleReLU<arma::mat, arma::mat>*,
Glimpse<arma::mat, arma::mat>*,
HardTanH<arma::mat, arma::mat>*,
Highway<arma::mat, arma::mat>*,
Join<arma::mat, arma::mat>*,
LayerNorm<arma::mat, arma::mat>*,
LeakyReLU<arma::mat, arma::mat>*,
CReLU<arma::mat, arma::mat>*,
Linear<arma::mat, arma::mat>*,
LinearNoBias<arma::mat, arma::mat>*,
Linear<arma::mat, arma::mat, NoRegularizer>*,
LinearNoBias<arma::mat, arma::mat, NoRegularizer>*,
LogSoftMax<arma::mat, arma::mat>*,
Lookup<arma::mat, arma::mat>*,
LSTM<arma::mat, arma::mat>*,
@@ -186,19 +236,13 @@ using LayerTypes = boost::variant<
FastLSTM<arma::mat, arma::mat>*,
MaxPooling<arma::mat, arma::mat>*,
MeanPooling<arma::mat, arma::mat>*,
MiniBatchDiscrimination<arma::mat, arma::mat>*,
MultiplyConstant<arma::mat, arma::mat>*,
MultiplyMerge<arma::mat, arma::mat>*,
NegativeLogLikelihood<arma::mat, arma::mat>*,
Padding<arma::mat, arma::mat>*,
PReLU<arma::mat, arma::mat>*,
Recurrent<arma::mat, arma::mat>*,
RecurrentAttention<arma::mat, arma::mat>*,
ReinforceNormal<arma::mat, arma::mat>*,
Reparametrization<arma::mat, arma::mat>*,
Select<arma::mat, arma::mat>*,
Sequential<arma::mat, arma::mat, false>*,
Sequential<arma::mat, arma::mat, true>*,
Subview<arma::mat, arma::mat>*,
VRClassReward<arma::mat, arma::mat>*,
MoreTypes,
CustomLayers*...
>;
src/mlpack/methods/ann/layer/linear.hpp
+12 -2
View File
@@ -14,6 +14,7 @@
#define MLPACK_METHODS_ANN_LAYER_LINEAR_HPP
#include <mlpack/prereqs.hpp>
#include <mlpack/methods/ann/regularizer/no_regularizer.hpp>
#include "layer_types.hpp"
@@ -31,7 +32,8 @@ namespace ann /** Artificial Neural Network. */ {
*/
template <
typename InputDataType = arma::mat,
typename OutputDataType = arma::mat
typename OutputDataType = arma::mat,
typename RegularizerType = NoRegularizer
>
class Linear
{
@@ -45,7 +47,9 @@ class Linear
* @param inSize The number of input units.
* @param outSize The number of output units.
*/
Linear(const size_t inSize, const size_t outSize);
Linear(const size_t inSize,
const size_t outSize,
RegularizerType regularizer = RegularizerType());
/*
* Reset the layer parameter.
@@ -113,6 +117,9 @@ class Linear
//! Modify the gradient.
OutputDataType& Gradient() { return gradient; }
//! Modify the bias weights of the layer.
arma::mat& Bias() { return bias; }
/**
* Serialize the layer
*/
@@ -146,6 +153,9 @@ class Linear
//! Locally-stored output parameter object.
OutputDataType outputParameter;
//! Locally-stored regularizer object.
RegularizerType regularizer;
}; // class Linear
} // namespace ann
src/mlpack/methods/ann/layer/linear_impl.hpp
+28 -16
View File
@@ -19,50 +19,60 @@
namespace mlpack {
namespace ann /** Artificial Neural Network. */ {
template<typename InputDataType, typename OutputDataType>
Linear<InputDataType, OutputDataType>::Linear()
template<typename InputDataType, typename OutputDataType,
typename RegularizerType>
Linear<InputDataType, OutputDataType, RegularizerType>::Linear() :
inSize(0),
outSize(0)
{
// Nothing to do here.
}
template <typename InputDataType, typename OutputDataType>
Linear<InputDataType, OutputDataType>::Linear(
template<typename InputDataType, typename OutputDataType,
typename RegularizerType>
Linear<InputDataType, OutputDataType, RegularizerType>::Linear(
const size_t inSize,
const size_t outSize) :
const size_t outSize,
RegularizerType regularizer) :
inSize(inSize),
outSize(outSize)
outSize(outSize),
regularizer(regularizer)
{
weights.set_size(outSize * inSize + outSize, 1);
}
template<typename InputDataType, typename OutputDataType>
void Linear<InputDataType, OutputDataType>::Reset()
template<typename InputDataType, typename OutputDataType,
typename RegularizerType>
void Linear<InputDataType, OutputDataType, RegularizerType>::Reset()
{
weight = arma::mat(weights.memptr(), outSize, inSize, false, false);
bias = arma::mat(weights.memptr() + weight.n_elem,
outSize, 1, false, false);
}
template<typename InputDataType, typename OutputDataType>
template<typename InputDataType, typename OutputDataType,
typename RegularizerType>
template<typename eT>
void Linear<InputDataType, OutputDataType>::Forward(
void Linear<InputDataType, OutputDataType, RegularizerType>::Forward(
const arma::Mat<eT>&& input, arma::Mat<eT>&& output)
{
output = weight * input;
output.each_col() += bias;
}
template<typename InputDataType, typename OutputDataType>
template<typename InputDataType, typename OutputDataType,
typename RegularizerType>
template<typename eT>
void Linear<InputDataType, OutputDataType>::Backward(
void Linear<InputDataType, OutputDataType, RegularizerType>::Backward(
const arma::Mat<eT>&& /* input */, arma::Mat<eT>&& gy, arma::Mat<eT>&& g)
{
g = weight.t() * gy;
}
template<typename InputDataType, typename OutputDataType>
template<typename InputDataType, typename OutputDataType,
typename RegularizerType>
template<typename eT>
void Linear<InputDataType, OutputDataType>::Gradient(
void Linear<InputDataType, OutputDataType, RegularizerType>::Gradient(
const arma::Mat<eT>&& input,
arma::Mat<eT>&& error,
arma::Mat<eT>&& gradient)
@@ -71,11 +81,13 @@ void Linear<InputDataType, OutputDataType>::Gradient(
error * input.t());
gradient.submat(weight.n_elem, 0, gradient.n_elem - 1, 0) =
arma::sum(error, 1);
regularizer.Evaluate(weights, gradient);
}
template<typename InputDataType, typename OutputDataType>
template<typename InputDataType, typename OutputDataType,
typename RegularizerType>
template<typename Archive>
void Linear<InputDataType, OutputDataType>::serialize(
void Linear<InputDataType, OutputDataType, RegularizerType>::serialize(
Archive& ar, const unsigned int /* version */)
{
ar & BOOST_SERIALIZATION_NVP(inSize);
src/mlpack/methods/ann/layer/linear_no_bias.hpp
+9 -2
View File
@@ -14,6 +14,7 @@
#define MLPACK_METHODS_ANN_LAYER_LINEAR_NO_BIAS_HPP
#include <mlpack/prereqs.hpp>
#include <mlpack/methods/ann/regularizer/no_regularizer.hpp>
#include "layer_types.hpp"
@@ -31,7 +32,8 @@ namespace ann /** Artificial Neural Network. */ {
*/
template <
typename InputDataType = arma::mat,
typename OutputDataType = arma::mat
typename OutputDataType = arma::mat,
typename RegularizerType = NoRegularizer
>
class LinearNoBias
{
@@ -44,7 +46,9 @@ class LinearNoBias
* @param inSize The number of input units.
* @param outSize The number of output units.
*/
LinearNoBias(const size_t inSize, const size_t outSize);
LinearNoBias(const size_t inSize,
const size_t outSize,
RegularizerType regularizer = RegularizerType());
/*
* Reset the layer parameter.
@@ -142,6 +146,9 @@ class LinearNoBias
//! Locally-stored output parameter object.
OutputDataType outputParameter;
//! Locally-stored regularizer object.
RegularizerType regularizer;
}; // class LinearNoBias
} // namespace ann
src/mlpack/methods/ann/layer/linear_no_bias_impl.hpp
+29 -16
View File
@@ -19,57 +19,70 @@
namespace mlpack {
namespace ann /** Artificial Neural Network. */ {
template<typename InputDataType, typename OutputDataType>
LinearNoBias<InputDataType, OutputDataType>::LinearNoBias()
template<typename InputDataType, typename OutputDataType,
typename RegularizerType>
LinearNoBias<InputDataType, OutputDataType, RegularizerType>::LinearNoBias() :
inSize(0),
outSize(0)
{
// Nothing to do here.
}
template<typename InputDataType, typename OutputDataType>
LinearNoBias<InputDataType, OutputDataType>::LinearNoBias(
const size_t inSize, const size_t outSize) :
template<typename InputDataType, typename OutputDataType,
typename RegularizerType>
LinearNoBias<InputDataType, OutputDataType, RegularizerType>::LinearNoBias(
const size_t inSize,
const size_t outSize,
RegularizerType regularizer) :
inSize(inSize),
outSize(outSize)
outSize(outSize),
regularizer(regularizer)
{
weights.set_size(outSize * inSize, 1);
}
template <typename InputDataType, typename OutputDataType>
void LinearNoBias<InputDataType, OutputDataType>::Reset()
template<typename InputDataType, typename OutputDataType,
typename RegularizerType>
void LinearNoBias<InputDataType, OutputDataType, RegularizerType>::Reset()
{
weight = arma::mat(weights.memptr(), outSize, inSize, false, false);
}
template<typename InputDataType, typename OutputDataType>
template<typename InputDataType, typename OutputDataType,
typename RegularizerType>
template<typename eT>
void LinearNoBias<InputDataType, OutputDataType>::Forward(
void LinearNoBias<InputDataType, OutputDataType, RegularizerType>::Forward(
const arma::Mat<eT>&& input, arma::Mat<eT>&& output)
{
output = weight * input;
}
template<typename InputDataType, typename OutputDataType>
template<typename InputDataType, typename OutputDataType,
typename RegularizerType>
template<typename eT>
void LinearNoBias<InputDataType, OutputDataType>::Backward(
void LinearNoBias<InputDataType, OutputDataType, RegularizerType>::Backward(
const arma::Mat<eT>&& /* input */, arma::Mat<eT>&& gy, arma::Mat<eT>&& g)
{
g = weight.t() * gy;
}
template<typename InputDataType, typename OutputDataType>
template<typename InputDataType, typename OutputDataType,
typename RegularizerType>
template<typename eT>
void LinearNoBias<InputDataType, OutputDataType>::Gradient(
void LinearNoBias<InputDataType, OutputDataType, RegularizerType>::Gradient(
const arma::Mat<eT>&& input,
arma::Mat<eT>&& error,
arma::Mat<eT>&& gradient)
{
gradient.submat(0, 0, weight.n_elem - 1, 0) = arma::vectorise(
error * input.t());
regularizer.Evaluate(weights, gradient);
}
template<typename InputDataType, typename OutputDataType>
template<typename InputDataType, typename OutputDataType,
typename RegularizerType>
template<typename Archive>
void LinearNoBias<InputDataType, OutputDataType>::serialize(
void LinearNoBias<InputDataType, OutputDataType, RegularizerType>::serialize(
Archive& ar, const unsigned int /* version */)
{
ar & BOOST_SERIALIZATION_NVP(inSize);
src/mlpack/methods/ann/layer/minibatch_discrimination.hpp
+190
View File
@@ -0,0 +1,190 @@
/**
* @file minibatch_discrimination.hpp
* @author Saksham Bansal
*
* Definition of the MiniBatchDiscrimination layer class.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_ANN_LAYER_MINIBATCH_DISCRIMINATION_HPP
#define MLPACK_METHODS_ANN_LAYER_MINIBATCH_DISCRIMINATION_HPP
#include <mlpack/prereqs.hpp>
#include "layer_types.hpp"
namespace mlpack {
namespace ann /** Artificial Neural Network. */ {
/**
* Implementation of the MiniBatchDiscrimination layer. MiniBatchDiscrimination
* is a layer of the discriminator that allows the discriminator to look at
* multiple data examples in combination and perform what is called as
* mini-batch discrimination.
* This helps prevent the collapse of the generator parameters to a setting
* where it emits the same point. This happens because normally a
* discriminator will process each example independently and there will be
* no mechanism to diversify the outputs of the generator.
*
* For more information, see the following.
*
* @code
* @article{Goodfellow2016,
* author = {Tim Salimans, Ian Goodfellow, Wojciech Zaremba, Vicki Cheung,
* Alec Radford, Xi Chen},
* title = {Improved Techniques for Training GANs},
* year = {2016},
* url = {https://arxiv.org/abs/1606.03498},
* }
* @endcode
*
* @tparam InputDataType Type of the input data (arma::colvec, arma::mat,
* arma::sp_mat or arma::cube).
* @tparam OutputDataType Type of the output data (arma::colvec, arma::mat,
* arma::sp_mat or arma::cube).
*/
template <
typename InputDataType = arma::mat,
typename OutputDataType = arma::mat
>
class MiniBatchDiscrimination
{
public:
//! Create the MiniBatchDiscrimination object.
MiniBatchDiscrimination();
/**
* Create the MiniBatchDiscrimination layer object using the specified
* number of units.
*
* @param inSize The number of input units.
* @param outSize The number of output units.
* @param features The number of features to compute for each dimension.
*/
MiniBatchDiscrimination(const size_t inSize,
const size_t outSize,
const size_t features);
/**
* Reset the layer parameter.
*/
void Reset();
/**
* Ordinary feed-forward pass of a neural network, evaluating the function
* f(x) by propagating the activity forward through f.
*
* @param input Input data used for evaluating the specified function.
* @param output Resulting output activation.
*/
template<typename eT>
void Forward(const arma::Mat<eT>&& input, arma::Mat<eT>&& output);
/**
* Ordinary feed-backward pass of a neural network, calculating the function
* f(x) by propagating x backwards through f. Using the results from the
* feed-forward pass.
*
* @param input The propagated input activation.
* @param gy The backpropagated error.
* @param g The calculated gradient.
*/
template<typename eT>
void Backward(const arma::Mat<eT>&& /* input */,
arma::Mat<eT>&& gy,
arma::Mat<eT>&& g);
/**
* Calculate the gradient using the output delta and the input activation.
*
* @param input The input parameter used for calculating the gradient.
* @param error The calculated error.
* @param gradient The calculated gradient.
*/
template<typename eT>
void Gradient(const arma::Mat<eT>&& input,
arma::Mat<eT>&& /* error */,
arma::Mat<eT>&& gradient);
//! Get the parameters.
OutputDataType const& Parameters() const { return weights; }
//! Modify the parameters.
OutputDataType& Parameters() { return weights; }
//! Get the input parameter.
InputDataType const& InputParameter() const { return inputParameter; }
//! Modify the input parameter.
InputDataType& InputParameter() { return inputParameter; }
//! Get the output parameter.
OutputDataType const& OutputParameter() const { return outputParameter; }
//! Modify the output parameter.
OutputDataType& OutputParameter() { return outputParameter; }
//! Get the delta.
OutputDataType const& Delta() const { return delta; }
//! Modify the delta.
OutputDataType& Delta() { return delta; }
//! Get the gradient.
OutputDataType const& Gradient() const { return gradient; }
//! Modify the gradient.
OutputDataType& Gradient() { return gradient; }
/**
* Serialize the layer.
*/
template<typename Archive>
void serialize(Archive& ar, const unsigned int /* version */);
private:
//! Locally-stored dimensions of weight.
size_t A, B, C;
//! Locally-stored input batch size.
size_t batchSize;
//! Locally-stored temporary features object.
arma::mat tempM;
//! Locally-stored weight object.
OutputDataType weights;
//! Locally-stored weight parameters.
OutputDataType weight;
//! Locally-stored features of input.
arma::cube M;
//! Locally-stored delta for features object.
arma::cube deltaM;
//! Locally-stored L1 distances between features.
arma::cube distances;
//! Locally-stored delta object.
OutputDataType delta;
//! Locally-stored temporary delta object.
OutputDataType deltaTemp;
//! Locally-stored gradient object.
OutputDataType gradient;
//! Locally-stored input parameter object.
InputDataType inputParameter;
//! Locally-stored output parameter object.
OutputDataType outputParameter;
}; // class MiniBatchDiscrimination
} // namespace ann
} // namespace mlpack
// Include implementation.
#include "minibatch_discrimination_impl.hpp"
#endif
src/mlpack/methods/ann/layer/minibatch_discrimination_impl.hpp
+147
View File
@@ -0,0 +1,147 @@
/**
* @file minibatch_discrimination_impl.hpp
* @author Saksham Bansal
*
* Implementation of the MiniBatchDiscrimination layer class.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_ANN_LAYER_MINIBATCH_DISCRIMINATION_IMPL_HPP
#define MLPACK_METHODS_ANN_LAYER_MINIBATCH_DISCRIMINATION_IMPL_HPP
// In case it hasn't yet been included.
#include "minibatch_discrimination.hpp"
namespace mlpack {
namespace ann /** Artificial Neural Network. */ {
template<typename InputDataType, typename OutputDataType>
MiniBatchDiscrimination<InputDataType, OutputDataType
>::MiniBatchDiscrimination() :
A(0),
B(0),
C(0),
batchSize(0)
{
// Nothing to do here.
}
template <typename InputDataType, typename OutputDataType>
MiniBatchDiscrimination<InputDataType, OutputDataType
>::MiniBatchDiscrimination(
const size_t inSize,
const size_t outSize,
const size_t features) :
A(inSize),
B(outSize - inSize),
C(features),
batchSize(0)
{
weights.set_size(A * B * C, 1);
}
template<typename InputDataType, typename OutputDataType>
void MiniBatchDiscrimination<InputDataType, OutputDataType>::Reset()
{
weight = arma::mat(weights.memptr(), B * C, A, false, false);
}
template<typename InputDataType, typename OutputDataType>
template<typename eT>
void MiniBatchDiscrimination<InputDataType, OutputDataType>::Forward(
const arma::Mat<eT>&& input, arma::Mat<eT>&& output)
{
batchSize = input.n_cols;
tempM = weight * input;
M = arma::cube(tempM.memptr(), B, C, batchSize, false, false);
distances.set_size(B, batchSize, batchSize);
output.set_size(B, batchSize);
for (size_t i = 0; i < M.n_slices; i++)
{
output.col(i).ones();
for (size_t j = 0; j < M.n_slices; j++)
{
if (j < i)
{
output.col(i) += distances.slice(j).col(i);
}
else if (i == j)
{
continue;
}
else
{
distances.slice(i).col(j) =
arma::exp(-arma::sum(abs(M.slice(i) - M.slice(j)), 1));
output.col(i) += distances.slice(i).col(j);
}
}
}
output = join_cols(input, output); // (A + B) x batchSize
}
template<typename InputDataType, typename OutputDataType>
template<typename eT>
void MiniBatchDiscrimination<InputDataType, OutputDataType>::Backward(
const arma::Mat<eT>&& /* input */, arma::Mat<eT>&& gy, arma::Mat<eT>&& g)
{
g = gy.head_rows(A);
arma::Mat<eT> gM = gy.tail_rows(B);
deltaM.zeros(B, C, batchSize);
for (size_t i = 0; i < M.n_slices; i++)
{
for (size_t j = 0; j < M.n_slices; j++)
{
if (i == j)
{
continue;
}
arma::mat t = arma::sign(M.slice(i) - M.slice(j));
t.each_col() %=
distances.slice(std::min(i, j)).col(std::max(i, j)) % gM.col(i);
deltaM.slice(i) -= t;
deltaM.slice(j) += t;
}
}
deltaTemp = arma::mat(deltaM.memptr(), B * C, batchSize, false, false);
g += weight.t() * deltaTemp;
}
template<typename InputDataType, typename OutputDataType>
template<typename eT>
void MiniBatchDiscrimination<InputDataType, OutputDataType>::Gradient(
const arma::Mat<eT>&& input,
arma::Mat<eT>&& /* error */,
arma::Mat<eT>&& gradient)
{
gradient = arma::vectorise(deltaTemp * input.t());
}
template<typename InputDataType, typename OutputDataType>
template<typename Archive>
void MiniBatchDiscrimination<InputDataType, OutputDataType>::serialize(
Archive& ar, const unsigned int /* version */)
{
ar & BOOST_SERIALIZATION_NVP(A);
ar & BOOST_SERIALIZATION_NVP(B);
ar & BOOST_SERIALIZATION_NVP(C);
// This is inefficient, but we have to allocate this memory so that
// WeightSetVisitor gets the right size.
if (Archive::is_loading::value)
{
weights.set_size(A * B * C, 1);
}
}
} // namespace ann
} // namespace mlpack
#endif
src/mlpack/methods/ann/layer/padding.hpp
+119
View File
@@ -0,0 +1,119 @@
/**
* @file padding.hpp
* @author Saksham Bansal
*
* Definition of the Padding class that pads the incoming data.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_ANN_LAYER_PADDING_HPP
#define MLPACK_METHODS_ANN_LAYER_PADDING_HPP
#include <mlpack/prereqs.hpp>
#include <mlpack/methods/ann/layer/layer_traits.hpp>
namespace mlpack {
namespace ann /** Artificial Neural Network. */ {
/**
* Implementation of the Padding module class. The Padding module applies a bias term
* to the incoming data.
*
* @tparam InputDataType Type of the input data (arma::colvec, arma::mat,
* arma::sp_mat or arma::cube).
* @tparam OutputDataType Type of the output data (arma::colvec, arma::mat,
* arma::sp_mat or arma::cube).
*/
template <
typename InputDataType = arma::mat,
typename OutputDataType = arma::mat
>
class Padding
{
public:
/**
* Create the Padding object using the specified number of output units.
*
* @param padWLeft Left padding width of the input.
* @param padWLeft Right padding width of the input.
* @param padHTop Top padding height of the input.
* @param padHBottom Bottom padding height of the input.
*/
Padding(const size_t padWLeft = 0,
const size_t padWRight = 0,
const size_t padHTop = 0,
const size_t padHBottom = 0);
/**
* Ordinary feed forward pass of a neural network, evaluating the function
* f(x) by propagating the activity forward through f.
*
* @param input Input data used for evaluating the specified function.
* @param output Resulting output activation.
*/
template<typename eT>
void Forward(const arma::Mat<eT>&& input, arma::Mat<eT>&& output);
/**
* Ordinary feed backward pass of a neural network, calculating the function
* f(x) by propagating x backwards trough f. Using the results from the feed
* forward pass.
*
* @param input The propagated input activation.
* @param gy The backpropagated error.
* @param g The calculated gradient.
*/
template<typename eT>
void Backward(const arma::Mat<eT>&& /* input */,
const arma::Mat<eT>&& gy,
arma::Mat<eT>&& g);
//! Get the output parameter.
OutputDataType const& OutputParameter() const { return outputParameter; }
//! Modify the output parameter.
OutputDataType& OutputParameter() { return outputParameter; }
//! Get the delta.
OutputDataType const& Delta() const { return delta; }
//! Modify the delta.
OutputDataType& Delta() { return delta; }
/**
* Serialize the layer.
*/
template<typename Archive>
void serialize(Archive& ar, const unsigned int /* version */);
private:
//! Locally-stored left padding width.
size_t padWLeft;
//! Locally-stored right padding width.
size_t padWRight;
//! Locally-stored top padding height.
size_t padHTop;
//! Locally-stored bottom padding height.
size_t padHBottom;
//! Locally-stored number of rows and columns of input.
size_t nRows, nCols;
//! Locally-stored delta object.
OutputDataType delta;
//! Locally-stored output parameter object.
OutputDataType outputParameter;
}; // class Padding
} // namespace ann
} // namespace mlpack
// Include implementation.
#include "padding_impl.hpp"
#endif
src/mlpack/methods/ann/layer/padding_impl.hpp
+76
View File
@@ -0,0 +1,76 @@
/**
* @file padding_impl.hpp
* @author Saksham Bansal
*
* Implementation of the Padding class that adds padding to the incoming
* data.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_ANN_LAYER_PADDING_IMPL_HPP
#define MLPACK_METHODS_ANN_LAYER_PADDING_IMPL_HPP
// In case it hasn't yet been included.
#include "padding.hpp"
namespace mlpack {
namespace ann /** Artificial Neural Network. */ {
template<typename InputDataType, typename OutputDataType>
Padding<InputDataType, OutputDataType>::Padding(
const size_t padWLeft,
const size_t padWRight,
const size_t padHTop,
const size_t padHBottom) :
padWLeft(padWLeft),
padWRight(padWRight),
padHTop(padHTop),
padHBottom(padHBottom),
nRows(0),
nCols(0)
{
// Nothing to do here.
}
template<typename InputDataType, typename OutputDataType>
template<typename eT>
void Padding<InputDataType, OutputDataType>::Forward(
const arma::Mat<eT>&& input, arma::Mat<eT>&& output)
{
nRows = input.n_rows;
nCols = input.n_cols;
output = arma::zeros(nRows + padWLeft + padWRight,
nCols + padHTop + padHBottom);
output.submat(padWLeft, padHTop, padWLeft + nRows - 1,
padHTop + nCols - 1) = input;
}
template<typename InputDataType, typename OutputDataType>
template<typename eT>
void Padding<InputDataType, OutputDataType>::Backward(
const arma::Mat<eT>&& /* input */,
const arma::Mat<eT>&& gy,
arma::Mat<eT>&& g)
{
g = gy.submat(padWLeft, padHTop, padWLeft + nRows - 1,
padHTop + nCols - 1);
}
template<typename InputDataType, typename OutputDataType>
template<typename Archive>
void Padding<InputDataType, OutputDataType>::serialize(
Archive& ar, const unsigned int /* version */)
{
ar & BOOST_SERIALIZATION_NVP(padWLeft);
ar & BOOST_SERIALIZATION_NVP(padWRight);
ar & BOOST_SERIALIZATION_NVP(padHTop);
ar & BOOST_SERIALIZATION_NVP(padHBottom);
}
} // namespace ann
} // namespace mlpack
#endif
src/mlpack/methods/ann/layer/transposed_convolution.hpp
+8 -48
View File
@@ -21,12 +21,13 @@
#include <mlpack/methods/ann/convolution_rules/svd_convolution.hpp>
#include "layer_types.hpp"
#include "padding.hpp"
namespace mlpack {
namespace ann /** Artificial Neural Network. */ {
/**
* Implementation of the Transposed Convolution class. The Transposed
* Implementation of the Transposed Convolution class. The Transposed
* Convolution class represents a single layer of a neural network.
*
* @tparam ForwardConvolutionRule Convolution to perform forward process.
@@ -162,6 +163,9 @@ class TransposedConvolution
//! Modify the output height.
size_t& OutputHeight() { return outputHeight; }
//! Modify the bias weights of the layer.
arma::mat& Bias() { return bias; }
/**
* Serialize the layer
*/
@@ -216,53 +220,6 @@ class TransposedConvolution
output = arma::fliplr(arma::flipud(input));
}
/*
* Pad the given input data.
*
* @param input The input to be padded.
* @param wPad Padding width of the input.
* @param hPad Padding height of the input.
* @param output The padded output data.
*/
template<typename eT>
void Pad(const arma::Mat<eT>& input,
size_t wPad,
size_t hPad,
arma::Mat<eT>& output)
{
if (output.n_rows != input.n_rows + wPad * 2 ||
output.n_cols != input.n_cols + hPad * 2)
{
output = arma::zeros(input.n_rows + wPad * 2, input.n_cols + hPad * 2);
}
output.submat(wPad, hPad, wPad + input.n_rows - 1,
hPad + input.n_cols - 1) = input;
}
/*
* Pad the given input data.
*
* @param input The input to be padded.
* @param wPad Padding width of the input.
* @param hPad Padding height of the input.
* @param output The padded output data.
*/
template<typename eT>
void Pad(const arma::Cube<eT>& input,
size_t wPad,
size_t hPad,
arma::Cube<eT>& output)
{
output = arma::zeros(input.n_rows + wPad * 2,
input.n_cols + hPad * 2, input.n_slices);
for (size_t i = 0; i < input.n_slices; ++i)
{
Pad<eT>(input.slice(i), wPad, hPad, output.slice(i));
}
}
//! Locally-stored number of input channels.
size_t inSize;
@@ -326,6 +283,9 @@ class TransposedConvolution
//! Locally-stored transformed gradient parameter.
arma::cube gradientTemp;
//! Locally-stored padding layer.
Padding<>* padding;
//! Locally-stored delta object.
OutputDataType delta;
src/mlpack/methods/ann/layer/transposed_convolution_impl.hpp
+1
View File
@@ -75,6 +75,7 @@ TransposedConvolution<
outputHeight(0)
{
weights.set_size((outSize * inSize * kW * kH) + outSize, 1);
padding = new Padding<>(padW, padW, padH, padH);
}
template<
src/mlpack/methods/ann/layer/virtual_batch_norm.hpp
+193
View File
@@ -0,0 +1,193 @@
/**
* @file virtual_batch_norm.hpp
* @author Saksham Bansal
*
* Definition of the VirtualBatchNorm layer class.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_ANN_LAYER_VIRTUALBATCHNORM_HPP
#define MLPACK_METHODS_ANN_LAYER_VIRTUALBATCHNORM_HPP
#include <mlpack/prereqs.hpp>
namespace mlpack {
namespace ann /** Artificial Neural Network. */ {
/**
* Declaration of the VirtualBatchNorm layer class. Instead of using the
* batch statistics for normalizing on a mini-batch, it uses a reference subset of
* the data for calculating the normalization statistics.
*
* For more information, refer to the following paper,
*
* @code
* @article{Goodfellow2016,
* author = {Tim Salimans, Ian Goodfellow, Wojciech Zaremba, Vicki Cheung,
* Alec Radford, Xi Chen},
* title = {Improved Techniques for Training GANs},
* year = {2016},
* url = {https://arxiv.org/abs/1606.03498},
* }
* @endcode
*
* @tparam InputDataType Type of the input data (arma::colvec, arma::mat,
* arma::sp_mat or arma::cube).
* @tparam OutputDataType Type of the output data (arma::colvec, arma::mat,
* arma::sp_mat or arma::cube).
*/
template <
typename InputDataType = arma::mat,
typename OutputDataType = arma::mat
>
class VirtualBatchNorm
{
public:
//! Create the VirtualBatchNorm object.
VirtualBatchNorm();
/**
* Create the VirtualBatchNorm layer object for a specified number of input units.
*
* @param referenceBatch The data from which the normalization
* statistics are computed.
* @param size The number of input units.
* @param eps The epsilon added to variance to ensure numerical stability.
*/
template<typename eT>
VirtualBatchNorm(const arma::Mat<eT>& referenceBatch,
const size_t size,
const double eps = 1e-8);
/**
* Reset the layer parameters.
*/
void Reset();
/**
* Forward pass of the Virtual Batch Normalization layer. Transforms the input data
* into zero mean and unit variance, scales the data by a factor gamma and
* shifts it by beta.
*
* @param input Input data for the layer.
* @param output Resulting output activations.
*/
template<typename eT>
void Forward(const arma::Mat<eT>&& input, arma::Mat<eT>&& output);
/**
* Backward pass through the layer.
*
* @param input The input activations.
* @param gy The backpropagated error.
* @param g The calculated gradient.
*/
template<typename eT>
void Backward(const arma::Mat<eT>&& /* input */,
arma::Mat<eT>&& gy,
arma::Mat<eT>&& g);
/**
* Calculate the gradient using the output delta and the input activations.
*
* @param input The input activations.
* @param error The calculated error.
* @param gradient The calculated gradient.
*/
template<typename eT>
void Gradient(const arma::Mat<eT>&& /* input */,
arma::Mat<eT>&& error,
arma::Mat<eT>&& gradient);
//! Get the parameters.
OutputDataType const& Parameters() const { return weights; }
//! Modify the parameters.
OutputDataType& Parameters() { return weights; }
//! Get the output parameter.
OutputDataType const& OutputParameter() const { return outputParameter; }
//! Modify the output parameter.
OutputDataType& OutputParameter() { return outputParameter; }
//! Get the delta.
OutputDataType const& Delta() const { return delta; }
//! Modify the delta.
OutputDataType& Delta() { return delta; }
//! Get the gradient.
OutputDataType const& Gradient() const { return gradient; }
//! Modify the gradient.
OutputDataType& Gradient() { return gradient; }
/**
* Serialize the layer.
*/
template<typename Archive>
void serialize(Archive& ar, const unsigned int /* version */);
private:
//! Locally-stored number of input units.
size_t size;
//! Locally-stored epsilon value.
double eps;
//! Variable to keep track of whether we are in loading or saving mode.
bool loading;
//! Locally-stored scale parameter.
OutputDataType gamma;
//! Locally-stored shift parameter.
OutputDataType beta;
//! Locally-stored parameters.
OutputDataType weights;
//! Mean of features in the reference batch.
OutputDataType referenceBatchMean;
//! Variance of features in the reference batch.
OutputDataType referenceBatchMeanSquared;
//! The coefficient for reference batch statistics.
double oldCoefficient;
//! The coefficient for input batch statistics.
double newCoefficient;
//! Locally-stored mean object.
OutputDataType mean;
//! Locally-stored variance object.
OutputDataType variance;
//! Locally-stored gradient object.
OutputDataType gradient;
//! Locally-stored delta object.
OutputDataType delta;
//! Locally-stored output parameter object.
OutputDataType outputParameter;
//! Locally-stored input parameter object.
OutputDataType inputParameter;
//! Locally-stored normalized input.
OutputDataType normalized;
//! Locally-stored zero mean input.
OutputDataType inputSubMean;
}; // class VirtualBatchNorm
} // namespace ann
} // namespace mlpack
// Include the implementation.
#include "virtual_batch_norm_impl.hpp"
#endif
src/mlpack/methods/ann/layer/virtual_batch_norm_impl.hpp
+153
View File
@@ -0,0 +1,153 @@
/**
* @file virtual_batch_norm_impl.hpp
* @author Saksham Bansal
*
* Implementation of the VirtualBatchNorm layer.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_ANN_LAYER_VIRTUALBATCHNORM_IMPL_HPP
#define MLPACK_METHODS_ANN_LAYER_VIRTUALBATCHNORM_IMPL_HPP
// In case it is not included.
#include "virtual_batch_norm.hpp"
namespace mlpack {
namespace ann { /** Artificial Neural Network. */
template<typename InputDataType, typename OutputDataType>
VirtualBatchNorm<InputDataType, OutputDataType>::VirtualBatchNorm() :
size(0),
eps(1e-8),
loading(false),
oldCoefficient(0),
newCoefficient(0)
{
// Nothing to do here.
}
template <typename InputDataType, typename OutputDataType>
template<typename eT>
VirtualBatchNorm<InputDataType, OutputDataType>::VirtualBatchNorm(
const arma::Mat<eT>& referenceBatch,
const size_t size,
const double eps) :
size(size),
eps(eps),
loading(false)
{
weights.set_size(size + size, 1);
referenceBatchMean = arma::mean(referenceBatch, 1);
referenceBatchMeanSquared = arma::mean(arma::square(referenceBatch), 1);
newCoefficient = 1.0 / (referenceBatch.n_cols + 1);
oldCoefficient = 1 - newCoefficient;
}
template<typename InputDataType, typename OutputDataType>
void VirtualBatchNorm<InputDataType, OutputDataType>::Reset()
{
gamma = arma::mat(weights.memptr(), size, 1, false, false);
beta = arma::mat(weights.memptr() + gamma.n_elem, size, 1, false, false);
if (!loading)
{
gamma.fill(1.0);
beta.fill(0.0);
}
loading = false;
}
template<typename InputDataType, typename OutputDataType>
template<typename eT>
void VirtualBatchNorm<InputDataType, OutputDataType>::Forward(
const arma::Mat<eT>&& input, arma::Mat<eT>&& output)
{
inputParameter = input;
arma::mat inputMean = arma::mean(input, 1);
arma::mat inputMeanSquared = arma::mean(arma::square(input), 1);
mean = oldCoefficient * referenceBatchMean + newCoefficient * inputMean;
arma::mat meanSquared = oldCoefficient * referenceBatchMeanSquared +
newCoefficient * inputMeanSquared;
variance = meanSquared - arma::square(mean);
// Normalize the input.
output = input.each_col() - mean;
inputSubMean = output;
output.each_col() /= arma::sqrt(variance + eps);
// Reused in the backward and gradient step.
normalized = output;
// Scale and shift the output.
output.each_col() %= gamma;
output.each_col() += beta;
}
template<typename InputDataType, typename OutputDataType>
template<typename eT>
void VirtualBatchNorm<InputDataType, OutputDataType>::Backward(
const arma::Mat<eT>&& /* input */, arma::Mat<eT>&& gy, arma::Mat<eT>&& g)
{
const arma::mat stdInv = 1.0 / arma::sqrt(variance + eps);
// dl / dxhat.
const arma::mat norm = gy.each_col() % gamma;
// sum dl / dxhat * (x - mu) * -0.5 * stdInv^3.
const arma::mat var = arma::sum(norm % inputSubMean, 1) %
arma::pow(stdInv, 3.0) * -0.5;
// dl / dxhat * 1 / stdInv + variance * 2 * (x - mu) / m +
// dl / dmu * newCoefficient / m.
g = (norm.each_col() % stdInv) + ((inputParameter.each_col() %
var) * 2 * newCoefficient / inputParameter.n_cols);
// (sum (dl / dxhat * -1 / stdInv) + (variance * mean * -2)) *
// newCoefficient / m.
g.each_col() += (arma::sum(norm.each_col() % -stdInv, 1) + (var %
mean * -2)) * newCoefficient / inputParameter.n_cols;
}
template<typename InputDataType, typename OutputDataType>
template<typename eT>
void VirtualBatchNorm<InputDataType, OutputDataType>::Gradient(
const arma::Mat<eT>&& /* input */,
arma::Mat<eT>&& error,
arma::Mat<eT>&& gradient)
{
gradient.set_size(size + size, 1);
// Step 5: dl / dy * xhat.
gradient.submat(0, 0, gamma.n_elem - 1, 0) = arma::sum(normalized % error, 1);
// Step 6: dl / dy.
gradient.submat(gamma.n_elem, 0, gradient.n_elem - 1, 0) =
arma::sum(error, 1);
}
template<typename InputDataType, typename OutputDataType>
template<typename Archive>
void VirtualBatchNorm<InputDataType, OutputDataType>::serialize(
Archive& ar, const unsigned int /* version */)
{
ar & BOOST_SERIALIZATION_NVP(size);
if (Archive::is_loading::value)
{
weights.set_size(size + size, 1);
loading = false;
}
ar & BOOST_SERIALIZATION_NVP(eps);
ar & BOOST_SERIALIZATION_NVP(gamma);
ar & BOOST_SERIALIZATION_NVP(beta);
}
} // namespace ann
} // namespace mlpack
#endif
src/mlpack/methods/ann/layer/weight_norm.hpp
+204
View File
@@ -0,0 +1,204 @@
/**
* @file weight_norm.hpp
* @author Toshal Agrawal
*
* Definition of the WeightNorm layer class.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_ANN_LAYER_WEIGHTNORM_HPP
#define MLPACK_METHODS_ANN_LAYER_WEIGHTNORM_HPP
#include <mlpack/prereqs.hpp>
#include "layer_types.hpp"
#include "../visitor/delete_visitor.hpp"
#include "../visitor/delta_visitor.hpp"
#include "../visitor/output_parameter_visitor.hpp"
#include "../visitor/reset_visitor.hpp"
#include "../visitor/weight_size_visitor.hpp"
#include "../visitor/weight_set_visitor.hpp"
namespace mlpack {
namespace ann /** Artificial Neural Network. */ {
/**
* Declaration of the WeightNorm layer class. The layer reparameterizes
* the weight vectors in a neural network, decoupling the length of those weight
* vectors from their direction. This reparameterization does not introduce any
* dependencies between the examples in a mini-batch.
*
* This class will be a wrapper around existing layers. It will just modify the
* calculation and updation of weights of the layer.
*
* For more information, refer to the following paper,
*
* @code
* @inproceedings{Salimans2016WeightNorm,
* title = {Weight Normalization: A Simple Reparameterization to Accelerate
* Training of Deep Neural Networks},
* author = {Tim Salimans, Diederik P. Kingma},
* booktitle = {Neural Information Processing Systems 2016},
* year = {2016}
* }
* @endcode
*
* @tparam InputDataType Type of the input data (arma::colvec, arma::mat,
* arma::sp_mat or arma::cube).
* @tparam OutputDataType Type of the output data (arma::colvec, arma::mat,
* arma::sp_mat or arma::cube).
* @tparam CustomLayers Additional custom layers that can be added.
*/
template <
typename InputDataType = arma::mat,
typename OutputDataType = arma::mat,
typename... CustomLayers
>
class WeightNorm
{
public:
/**
* Create the WeightNorm layer object.
*
* @param layer The layer whose weights are needed to be normalized.
*/
WeightNorm(LayerTypes<CustomLayers...> layer = LayerTypes<CustomLayers...>());
//! Destructor to release allocated memory.
~WeightNorm();
/**
* Reset the layer parameters.
*/
void Reset();
/**
* Forward pass of the WeightNorm layer. Calculates the weights of the
* wrapped layer from the parameter vector v and the scalar parameter g.
* It then calulates the output of the wrapped layer from the calculated
* weights.
*
* @param input Input data for the layer.
* @param output Resulting output activations.
*/
template<typename eT>
void Forward(arma::Mat<eT>&& input, arma::Mat<eT>&& output);
/**
* Backward pass through the layer. This function calls the Backward()
* function of the wrapped layer.
*
* @param input The input activations.
* @param gy The backpropagated error.
* @param g The calculated gradient.
*/
template<typename eT>
void Backward(const arma::Mat<eT>&& input,
arma::Mat<eT>&& gy,
arma::Mat<eT>&& g);
/**
* Calculate the gradient using the output delta, input activations and the
* weights of the wrapped layer.
*
* @param input The input activations.
* @param error The calculated error.
* @param gradient The calculated gradient.
*/
template<typename eT>
void Gradient(arma::Mat<eT>&& input,
arma::Mat<eT>&& error,
arma::Mat<eT>&& gradient);
//! Get the delta.
OutputDataType const& Delta() const { return delta; }
//! Modify the delta.
OutputDataType& Delta() { return delta; }
//! Get the gradient.
OutputDataType const& Gradient() const { return gradient; }
//! Modify the gradient.
OutputDataType& Gradient() { return gradient; }
//! Get the output parameter.
OutputDataType const& OutputParameter() const { return outputParameter; }
//! Modify the output parameter.
OutputDataType& OutputParameter() { return outputParameter; }
//! Get the parameters.
OutputDataType const& Parameters() const { return weights; }
//! Modify the parameters.
OutputDataType& Parameters() { return weights; }
//! Get the wrapped layer.
LayerTypes<CustomLayers...> const& Layer() { return wrappedLayer; }
/**
* Serialize the layer.
*/
template<typename Archive>
void serialize(Archive& ar, const unsigned int /* version */);
private:
//! Locally-stored number of bias elements in the weights of wrapped layer.
size_t biasWeightSize;
//! Locally-stored delete visitor module object.
DeleteVisitor deleteVisitor;
//! Locally-stored delta object.
OutputDataType delta;
//! Locally-stored delta visitor module object.
DeltaVisitor deltaVisitor;
//! Locally-stored gradient object.
OutputDataType gradient;
//! Locally-stored wrapped layer.
LayerTypes<CustomLayers...> wrappedLayer;
//! Locally stored number of elements in the weights of wrapped layer.
size_t layerWeightSize;
//! Locally-stored output parameter object.
OutputDataType outputParameter;
//! Locally-stored output parameter visitor module object.
OutputParameterVisitor outputParameterVisitor;
//! Reset the gradient for all modules that implement the Gradient function.
void ResetGradients(arma::mat& gradient);
//! Locally-stored reset visitor.
ResetVisitor resetVisitor;
//! Locally-stored scalar parameter.
OutputDataType scalarParameter;
//! Locally-stored parameter vector.
OutputDataType vectorParameter;
//! Locally-stored parameters.
OutputDataType weights;
//! Locally-stored weight size visitor.
WeightSizeVisitor weightSizeVisitor;
//! Locally-stored gradients of wrappedLayer.
OutputDataType layerGradients;
//! Locally-stored weights of wrappedLayer.
OutputDataType layerWeights;
}; // class WeightNorm
} // namespace ann
} // namespace mlpack
// Include the implementation.
#include "weight_norm_impl.hpp"
#endif
src/mlpack/methods/ann/layer/weight_norm_impl.hpp
+168
View File
@@ -0,0 +1,168 @@
/**
* @file weight_norm_impl.hpp
* @author Toshal Agrawal
*
* Implementation of the WeightNorm Layer.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_ANN_LAYER_WEIGHTNORM_IMPL_HPP
#define MLPACK_METHODS_ANN_LAYER_WEIGHTNORM_IMPL_HPP
// In case it is not included.
#include "weight_norm.hpp"
#include "../visitor/forward_visitor.hpp"
#include "../visitor/backward_visitor.hpp"
#include "../visitor/gradient_visitor.hpp"
#include "../visitor/bias_set_visitor.hpp"
namespace mlpack {
namespace ann { /** Artificial Neural Network. */
template<typename InputDataType, typename OutputDataType,
typename... CustomLayers>
WeightNorm<InputDataType, OutputDataType, CustomLayers...>::WeightNorm(
LayerTypes<CustomLayers...> layer) :
wrappedLayer(layer)
{
layerWeightSize = boost::apply_visitor(weightSizeVisitor, wrappedLayer);
weights.set_size(layerWeightSize + 1, 1);
layerWeights.set_size(layerWeightSize, 1);
layerGradients.set_size(layerWeightSize, 1);
}
template<typename InputDataType, typename OutputDataType,
typename... CustomLayers>
WeightNorm<InputDataType, OutputDataType, CustomLayers...>::~WeightNorm()
{
boost::apply_visitor(deleteVisitor, wrappedLayer);
}
template<typename InputDataType, typename OutputDataType,
typename... CustomLayers>
void WeightNorm<InputDataType, OutputDataType, CustomLayers...>::Reset()
{
// Set the weights of the inside layer to layerWeights.
// This is done to set the non-bias terms correctly.
boost::apply_visitor(WeightSetVisitor(std::move(layerWeights), 0),
wrappedLayer);
boost::apply_visitor(resetVisitor, wrappedLayer);
biasWeightSize = boost::apply_visitor(BiasSetVisitor(std::move(weights),
0), wrappedLayer);
vectorParameter = arma::mat(weights.memptr() + biasWeightSize,
layerWeightSize - biasWeightSize, 1, false, false);
scalarParameter = arma::mat(weights.memptr() + layerWeightSize, 1, 1, false,
false);
}
template<typename InputDataType, typename OutputDataType,
typename... CustomLayers>
template<typename eT>
void WeightNorm<InputDataType, OutputDataType, CustomLayers...>::Forward(
arma::Mat<eT>&& input, arma::Mat<eT>&& output)
{
// Initialize the non-bias weights of wrapped layer.
const double normVectorParameter = arma::norm(vectorParameter, 2);
layerWeights.rows(0, layerWeightSize - biasWeightSize - 1) =
scalarParameter(0) * vectorParameter / normVectorParameter;
boost::apply_visitor(ForwardVisitor(std::move(input), std::move(
boost::apply_visitor(outputParameterVisitor, wrappedLayer))),
wrappedLayer);
output = boost::apply_visitor(outputParameterVisitor, wrappedLayer);
}
template<typename InputDataType, typename OutputDataType,
typename... CustomLayers>
template<typename eT>
void WeightNorm<InputDataType, OutputDataType, CustomLayers...>::Backward(
const arma::Mat<eT>&& /* input */, arma::Mat<eT>&& gy, arma::Mat<eT>&& g)
{
boost::apply_visitor(BackwardVisitor(std::move(boost::apply_visitor(
outputParameterVisitor, wrappedLayer)), std::move(gy), std::move(
boost::apply_visitor(deltaVisitor, wrappedLayer))), wrappedLayer);
g = boost::apply_visitor(deltaVisitor, wrappedLayer);
}
template<typename InputDataType, typename OutputDataType,
typename... CustomLayers>
template<typename eT>
void WeightNorm<InputDataType, OutputDataType, CustomLayers...>::Gradient(
arma::Mat<eT>&& input,
arma::Mat<eT>&& error,
arma::Mat<eT>&& gradient)
{
ResetGradients(layerGradients);
// Calculate the gradients of the wrapped layer.
boost::apply_visitor(GradientVisitor(std::move(input),
std::move(error)), wrappedLayer);
// Store the norm of vector parameter temporarily.
const double normVectorParameter = arma::norm(vectorParameter, 2);
// Set the gradients of the bias terms.
if (biasWeightSize != 0)
{
gradient.rows(0, biasWeightSize - 1) = arma::mat(layerGradients.memptr() +
layerWeightSize - biasWeightSize, biasWeightSize, 1, false, false);
}
// Calculate the gradients of the scalar parameter.
gradient[gradient.n_rows - 1] = arma::accu(layerGradients.rows(0,
layerWeightSize - biasWeightSize - 1) % vectorParameter) /
normVectorParameter;
// Calculate the gradients of the vector parameter.
gradient.rows(biasWeightSize, layerWeightSize - 1) =
scalarParameter(0) / normVectorParameter * (layerGradients.rows(0,
layerWeightSize - biasWeightSize - 1) - gradient[gradient.n_rows - 1] /
normVectorParameter * vectorParameter);
}
template<typename InputDataType, typename OutputDataType,
typename... CustomLayers>
void WeightNorm<InputDataType, OutputDataType, CustomLayers...>::ResetGradients(
arma::mat& gradient)
{
boost::apply_visitor(GradientSetVisitor(std::move(gradient), 0),
wrappedLayer);
}
template<typename InputDataType, typename OutputDataType,
typename... CustomLayers>
template<typename Archive>
void WeightNorm<InputDataType, OutputDataType, CustomLayers...>::serialize(
Archive& ar, const unsigned int /* version */)
{
if (Archive::is_loading::value)
{
boost::apply_visitor(deleteVisitor, wrappedLayer);
}
ar & BOOST_SERIALIZATION_NVP(wrappedLayer);
ar & BOOST_SERIALIZATION_NVP(layerWeightSize);
// If we are loading, we need to initialize the weights.
if (Archive::is_loading::value)
{
weights.set_size(layerWeightSize + 1, 1);
}
}
} // namespace ann
} // namespace mlpack
#endif
src/mlpack/methods/ann/regularizer/CMakeLists.txt
+18
View File
@@ -0,0 +1,18 @@
# Define the files we need to compile.
# Anything not in this list will not be compiled into mlpack.
set(SOURCES
lregularizer.hpp
lregularizer_impl.hpp
orthogonal_regularizer.hpp
orthogonal_regularizer_impl.hpp
no_regularizer.hpp
)
# Add directory name to sources.
set(DIR_SRCS)
foreach(file ${SOURCES})
set(DIR_SRCS ${DIR_SRCS} ${CMAKE_CURRENT_SOURCE_DIR}/${file})
endforeach()
# Append sources (with directory name) to list of all mlpack sources (used at
# the parent scope).
set(MLPACK_SRCS ${MLPACK_SRCS} ${DIR_SRCS} PARENT_SCOPE)
src/mlpack/methods/ann/regularizer/lregularizer.hpp
+75
View File
@@ -0,0 +1,75 @@
/**
* @file lregularizer.hpp
* @author Saksham Bansal
*
* Generalized L-regularizer, allowing both l1 and l2 regularization methods.
* This also gives several convenience typedefs for commonly used L-regularizers.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_ANN_LREGULARIZER_HPP
#define MLPACK_METHODS_ANN_LREGULARIZER_HPP
#include <mlpack/prereqs.hpp>
namespace mlpack {
namespace ann {
/**
* The L_p regularizer for arbitrary integer p.
*
* @tparam Power Power of regularizer; i.e. Power = 1 gives the L1-regularization.
*/
template<int TPower>
class LRegularizer
{
public:
/**
* Create the regularizer object.
*
* @param factor The factor for regularization.
*/
LRegularizer(double factor = 1.0);
/**
* Calculate the gradient for regularization.
*
* @tparam MatType Type of weight matrix.
* @param weight The weight matrix to be regularized.
* @param gradient The calculated gradient.
*/
template<typename MatType>
void Evaluate(const MatType& weight, MatType& gradient);
//! Serialize the regularizer (nothing to do).
template<typename Archive>
void serialize(Archive& ar, const unsigned int /* version */);
//! The power of the regularizer.
static const int Power = TPower;
//! The constant for the regularization
double factor;
};
// Convenience typedefs.
/**
* The L1 Regularizer.
*/
typedef LRegularizer<1> L1Regularizer;
/**
* The L2 Regularizer.
*/
typedef LRegularizer<2> L2Regularizer;
} // namespace ann
} // namespace mlpack
// Include implementation.
#include "lregularizer_impl.hpp"
#endif
src/mlpack/methods/ann/regularizer/lregularizer_impl.hpp
+63
View File
@@ -0,0 +1,63 @@
/**
* @file lregularization_impl.hpp
* @author Saksham Bansal
*
* Implementation of template specializations of LRegularizer class.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_ANN_LREGULARIZER_IMPL_HPP
#define MLPACK_METHODS_ANN_LREGULARIZER_IMPL_HPP
// In case it hasn't been included.
#include "lregularizer.hpp"
namespace mlpack {
namespace ann {
template<int Power>
LRegularizer<Power>::LRegularizer(double factor) :
factor(factor)
{
// Nothing to do here
}
// Unspecialized implementation. This should almost never be used...
template<int Power>
template<typename MatType>
void LRegularizer<Power>::Evaluate(const MatType& weight, MatType& gradient)
{
gradient += arma::vectorise(arma::pow(weight, Power - 1) * Power * factor);
}
// L1-Regularizer specializations.
template<>
template<typename MatType>
void LRegularizer<1>::Evaluate(const MatType& weight, MatType& gradient)
{
gradient += arma::vectorise(factor * weight / arma::abs(weight));
}
// L2-Regularizer specializations.
template<>
template<typename MatType>
void LRegularizer<2>::Evaluate(const MatType& weight, MatType& gradient)
{
gradient += arma::vectorise(2 * factor * weight);
}
template<int Power>
template<typename Archive>
void LRegularizer<Power>::serialize(
Archive& ar, const unsigned int /* version */)
{
ar & BOOST_SERIALIZATION_NVP(factor);
}
} // namespace ann
} // namespace mlpack
#endif
src/mlpack/methods/ann/regularizer/no_regularizer.hpp
+52
View File
@@ -0,0 +1,52 @@
/**
* @file no_regularizer.hpp
* @author Saksham Bansal
*
* Definition of the NoRegularizer class.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_ANN_NO_REGULARIZER_HPP
#define MLPACK_METHODS_ANN_NO_REGULARIZER_HPP
#include <mlpack/prereqs.hpp>
namespace mlpack {
namespace ann {
/**
* Implementation of the NoRegularizer. This does not add any
* regularization to the weights.
*/
class NoRegularizer
{
public:
/**
* Create the regularizer object.
*/
NoRegularizer()
{
// Nothing to do here.
};
/**
* Calculate the gradient for regularization.
*
* @tparam MatType Type of weight matrix.
* @param weight The weight matrix to be regularized.
* @param gradient The calculated gradient.
*/
template<typename MatType>
void Evaluate(const MatType& /* weight */, MatType& /* gradient */)
{
// Nothing to do here.
}
};
} // namespace ann
} // namespace mlpack
#endif
src/mlpack/methods/ann/regularizer/orthogonal_regularizer.hpp
+72
View File
@@ -0,0 +1,72 @@
/**
* @file orthogonal_regularizer.hpp
* @author Saksham Bansal
*
* Definition of the OrthogonalRegularizer class.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_ANN_ORTHOGONAL_REGULARIZER_HPP
#define MLPACK_METHODS_ANN_ORTHOGONAL_REGULARIZER_HPP
#include <mlpack/prereqs.hpp>
namespace mlpack {
namespace ann {
/**
* Implementation of the OrthogonalRegularizer. Orthogonality of weights is a
* desirable property because multiplication by an orthogonal matrix leaves
* the norm of the matrix unchanged. The orthogonal regularization technique
* encourages weights to be orthogonal.
*
* For more information, see the following.
*
* @code
* @inproceedings{WanICML2013,
* title={Neural Photo Editing with Introspective Adversarial Networks},
* booktitle = {5th International Conference on Learning Representations
* (ICLR - 17)},
* author = {Andrew Brock and Theodore Lim and J.M. Ritchie and Nick Weston},
* year = {2017}
* }
* @endcode
*/
class OrthogonalRegularizer
{
public:
/**
* Create the regularizer object.
*
* @param factor The factor for regularization.
*/
OrthogonalRegularizer(double factor = 1.0);
/**
* Calculate the gradient for regularization.
*
* @tparam MatType Type of weight matrix.
* @param weight The weight matrix to be regularized.
* @param gradient The calculated gradient.
*/
template<typename MatType>
void Evaluate(const MatType& weight, MatType& gradient);
//! Serialize the regularizer (nothing to do).
template<typename Archive>
void serialize(Archive& ar, const unsigned int /* version */);
//! The constant for the regularization
double factor;
};
} // namespace ann
} // namespace mlpack
// Include implementation.
#include "orthogonal_regularizer_impl.hpp"
#endif
src/mlpack/methods/ann/regularizer/orthogonal_regularizer_impl.hpp
+66
View File
@@ -0,0 +1,66 @@
/**
* @file orthogonal_regularizer_impl.hpp
* @author Saksham Bansal
*
* Implementation of OrthogonalRegularizer class.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_ANN_ORTHOGONAL_REGULARIZER_IMPL_HPP
#define MLPACK_METHODS_ANN_ORTHOGONAL_REGULARIZER_IMPL_HPP
// In case it hasn't been included.
#include "orthogonal_regularizer.hpp"
namespace mlpack {
namespace ann {
OrthogonalRegularizer::OrthogonalRegularizer(double factor) :
factor(factor)
{
// Nothing to do here.
}
template<typename MatType>
void OrthogonalRegularizer::Evaluate(const MatType& weight, MatType& gradient)
{
arma::mat grad = arma::zeros(arma::size(weight));
for (size_t i = 0; i < weight.n_rows; i++)
{
for (size_t j = 0; j < weight.n_rows; j++)
{
if (i == j)
{
double s =
arma::as_scalar(
arma::sign((weight.row(i) * weight.row(i).t()) - 1));
grad.row(i) += 2 * s * weight.row(i);
}
else
{
double s = arma::as_scalar(
arma::sign(weight.row(i) * weight.row(j).t()));
grad.row(i) += s * weight.row(j);
grad.row(j) += s * weight.row(i);
}
}
}
gradient += arma::vectorise(grad) * factor;
}
template<typename Archive>
void OrthogonalRegularizer::serialize(
Archive& ar, const unsigned int /* version */)
{
ar & BOOST_SERIALIZATION_NVP(factor);
}
} // namespace ann
} // namespace mlpack
#endif
src/mlpack/methods/ann/regularizer/regularizer.hpp
+19
View File
@@ -0,0 +1,19 @@
/**
* @file regularizer.hpp
* @author Saksham Bansal
*
* This includes various regularizers to construct a model.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_ANN_REGULARIZER_HPP
#define MLPACK_METHODS_ANN_REGULARIZER_HPP
#include "lregularizer.hpp"
#include "orthogonal_regularizer.hpp"
#include "no_regularizer.hpp"
#endif
src/mlpack/methods/ann/visitor/CMakeLists.txt
+2
View File
@@ -5,6 +5,8 @@ set(SOURCES
add_visitor_impl.hpp
backward_visitor.hpp
backward_visitor_impl.hpp
bias_set_visitor.hpp
bias_set_visitor_impl.hpp
copy_visitor.hpp
copy_visitor_impl.hpp
delete_visitor.hpp
src/mlpack/methods/ann/visitor/add_visitor.hpp
+2
View File
@@ -36,6 +36,8 @@ class AddVisitor : public boost::static_visitor<void>
template<typename LayerType>
void operator()(LayerType* layer) const;
void operator()(MoreTypes layer) const;
private:
//! The layer that should be added.
LayerTypes<CustomLayers...> newLayer;
src/mlpack/methods/ann/visitor/add_visitor_impl.hpp
+6
View File
@@ -34,6 +34,12 @@ inline void AddVisitor<CustomLayers...>::operator()(LayerType* layer) const
LayerAdd<LayerType>(layer);
}
template<typename... CustomLayers>
inline void AddVisitor<CustomLayers...>::operator()(MoreTypes layer) const
{
layer.apply_visitor(*this);
}
template<typename... CustomLayers>
template<typename T>
inline typename std::enable_if<
src/mlpack/methods/ann/visitor/backward_visitor.hpp
+2
View File
@@ -40,6 +40,8 @@ class BackwardVisitor : public boost::static_visitor<void>
template<typename LayerType>
void operator()(LayerType* layer) const;
void operator()(MoreTypes layer) const;
private:
//! The input parameter set.
arma::mat&& input;
src/mlpack/methods/ann/visitor/backward_visitor_impl.hpp
+5
View File
@@ -50,6 +50,11 @@ inline void BackwardVisitor::operator()(LayerType* layer) const
LayerBackward(layer, layer->OutputParameter());
}
inline void BackwardVisitor::operator()(MoreTypes layer) const
{
layer.apply_visitor(*this);
}
template<typename T>
inline typename std::enable_if<
!HasRunCheck<T, bool&(T::*)(void)>::value, void>::type
src/mlpack/methods/ann/visitor/bias_set_visitor.hpp
+82
View File
@@ -0,0 +1,82 @@
/**
* @file bias_set_visitor.hpp
* @author Toshal Agrawal
*
* This file provides an abstraction for the Bias() function for different
* layers and automatically directs any parameter to the right layer type.
*
* mlpack is free software; you may redistribute it and/or modify it under the
* terms of the 3-clause BSD license. You should have received a copy of the
* 3-clause BSD license along with mlpack. If not, see
* http://www.opensource.org/licenses/BSD-3-Clause for more information.
*/
#ifndef MLPACK_METHODS_ANN_VISITOR_BIAS_SET_VISITOR_HPP
#define MLPACK_METHODS_ANN_VISITOR_BIAS_SET_VISITOR_HPP
#include <mlpack/methods/ann/layer/layer_traits.hpp>
#include <boost/variant.hpp>
namespace mlpack {
namespace ann {
/**
* BiasSetVisitor updates the module bias parameters given the parameters set.
*/
class BiasSetVisitor : public boost::static_visitor<size_t>
{
public:
//! Update the bias parameters given the parameters' set and offset.
BiasSetVisitor(arma::mat&& weight, const size_t offset = 0);
//! Update the parameters' set.
template<typename LayerType>
size_t operator()(LayerType* layer) const;
size_t operator()(MoreTypes layer) const;
private:
//! The parameters' set.
arma::mat&& weight;
//! The parameters' offset.
const size_t offset;
//! Do not update the bias parameters if the module doesn't implement the
//! Bias() or Model() function.
template<typename T>
typename std::enable_if<
!HasBiasCheck<T, arma::mat&(T::*)()>::value &&
!HasModelCheck<T>::value, size_t>::type
LayerSize(T* layer) const;
//! Update the bias parameters if the module implements the Model() function.
template<typename T>
typename std::enable_if<
!HasBiasCheck<T, arma::mat&(T::*)()>::value &&
HasModelCheck<T>::value, size_t>::type
LayerSize(T* layer) const;
//! Update the bias parameters if the module implements the Bias() function.
template<typename T>
typename std::enable_if<
HasBiasCheck<T, arma::mat&(T::*)()>::value &&
!HasModelCheck<T>::value, size_t>::type
LayerSize(T* layer) const;
//! Update the bias parameters if the module implements the Model() and
//! Bias() function.
template<typename T>
typename std::enable_if<
HasBiasCheck<T, arma::mat&(T::*)()>::value &&
HasModelCheck<T>::value, size_t>::type
LayerSize(T* layer) const;
};
} // namespace ann
} // namespace mlpack
// Include implementation.
#include "bias_set_visitor_impl.hpp"
#endif

Some files were not shown because too many files have changed in this diff Show More