Commit Graph
12802 Commits
Author SHA1 Message Date
Ryan Curtin 3e4a3c8ba4 Note that softmax regression is done. 2015-03-05 11:23:36 -05:00
Ryan Curtin 251fb8f1a2 Change API slightly to match other code.
Realistically, this probably needs further thought: C++11 gives us rvalue references, so we can do something a bit smarter to validate parameters without any unnecessary copies that would result from Mutator(const T&) (or whatever).
2015-03-05 11:07:10 -05:00
Ryan Curtin f529c5036c Trivial spelling fix (why am I even bothering?). 2015-03-05 10:56:08 -05:00
Ryan Curtin 1d19aa1e1a Add new contributor. 2015-03-05 10:51:57 -05:00
Ryan Curtin 672d8f0acb Merge pull request #418 from apir8181/issue-332
issue332 - add intercept term to softmax regression
2015-03-05 10:50:36 -05:00
apir8181 918df7156c add intercept term to softmax regression 2015-03-05 16:01:38 +08:00
Ryan Curtin a62991ace0 Add new contributors. 2015-03-03 15:46:01 -05:00
Ryan Curtin 3642663285 Fix documentation, pointed out by Marcus.
Also fix the return type when we can't transpose.
2015-03-03 15:43:16 -05:00
Ryan Curtin 0a7c4f4651 Armadillo 3.930+ supports inplace_trans() too. 2015-03-03 11:31:31 -05:00
Stephen Tu df09588fc1 arma::inplace_trans is not available for older versions of arma 2015-03-02 19:56:07 -08:00
Ryan Curtin 534c7d357d Remove references to Boost.Random. 2015-03-02 20:45:35 -05:00
Ryan Curtin bca4027b27 Merge pull request #410 from VladimirGl/randoms
Use c++11 random.
2015-03-02 20:42:31 -05:00
Ryan Curtin b7769dcffb I've concluded that this test isn't useful.
Basically, the test is just checking that the implementation is exactly the same
as in the test, and that's not actually a great test, because the implementation
may change without breaking anything.
2015-03-02 20:24:04 -05:00
Ryan Curtin 2b76e6d0ac This isn't warning output. Make it "info" instead. 2015-03-02 11:24:21 -05:00
ryan 1703198190 Ensmallen dataset; test speedup about 25x.
The cost may be the possibility of a bug going unnoticed, but given that this is a random test anyway, we're basically just slightly increasing P(fastmks is wrong | test passes), but that should be quite small anyway.
2015-03-01 17:33:09 -05:00
Marcus Edel f5c8e3c97d Minor style changes. 2015-02-28 23:53:19 +01:00
Ryan Curtin fb5fec3caa Merge pull request #411 from VladimirGl/master
Added vim temp files to gitignore.
2015-02-27 16:49:33 -05:00
Vladimir Glazachev bd36cd0309 added vim temp files to gitignore 2015-02-28 01:42:50 +04:00
Marcus Edel 196cd879f7 Merge branch from github.com:zoq/mlpack to address structure issues. 2015-02-27 21:39:10 +01:00
Marcus Edel 923332420c Add test cases for the recurrent neural network. 2015-02-27 21:13:11 +01:00
Marcus Edel e933c9bce9 Properly handle the rnn structure. 2015-02-27 21:11:59 +01:00
Marcus Edel 22c413ff07 Use the correct number of epochs. 2015-02-27 21:10:39 +01:00
Marcus Edel 924da7bb42 Add function to evaluate the network. 2015-02-27 21:09:28 +01:00
Marcus Edel 55f9309562 Minor style fixes. 2015-02-27 21:08:38 +01:00
Ryan Curtin decc21edd6 Merge pull request #408 from VladimirGl/Issue#203
Adds inplace transpose method and some tests.
2015-02-27 13:58:44 -05:00
Vladimir Glazachev 35bd45ca1f use std random module instead of boost 2015-02-27 22:42:20 +04:00
Ryan Curtin 869bb06dcc Non-important spelling and style fixes. 2015-02-24 11:18:36 -05:00
Ryan Curtin 5f26cb0bf2 Merge pull request #409 from apir8181/issue-344
Issue 344 - Better boost test output
2015-02-24 11:16:14 -05:00
apir8181 581667c184 better boost test output 2015-02-24 23:47:58 +08:00
Vladimir Glazachev 73d5ff9535 inpace transpoce moved to load_impl.cpp 2015-02-24 01:57:27 +04:00
Vladimir Glazachev 7881c3e3e6 Adds inplace transpose method and some tests. This method used in Load() now. 2015-02-22 04:51:43 +04:00
Ryan Curtin 23c21da422 Merge pull request #407 from hritik25/patch-1
Update TREE_EXPLANATION.txt
2015-02-20 23:38:19 -05:00
hritik25 d5f5b9b3cd Update TREE_EXPLANATION.txt
Small typo
2015-02-21 03:46:41 +05:30
Ryan Curtin ce93257e6a Force high-precision Armadillo saves.
This will fix the failing tests in GMMTest and HMMTest.
We now use enable_if_c<> to catch when an Armadillo dense or sparse object is
being saved/loaded, and use Armadillo's diskio functionality to write to (or
read from) a stream.  This allows high-precision saving, whereas just printing
via `stream << mat` is pretty low-precision and can't be changed with
setwidth().
2015-02-16 15:46:01 -05:00
Stephen Tu 817c8b5a61 travis.yml: don't use travis_wait 2015-02-15 12:10:58 -08:00
Stephen Tu d091d37ccf travis.yml: try travis_wait 2015-02-15 11:06:27 -08:00
Stephen Tu 986736fa72 travis.yml: log level all for tests 2015-02-15 10:50:51 -08:00
Stephen Tu 15fdcedec0 travis.yml: pwd for debug 2015-02-14 23:21:33 -08:00
Stephen Tu 63359676b4 travis.yml: make test 2015-02-14 23:07:22 -08:00
Stephen Tu 59beaf142a travis: try ARMA_64BIT_WORD 2015-02-14 22:58:39 -08:00
Stephen Tu 79150123e8 travis.yml: i hate you 2015-02-14 22:49:26 -08:00
Stephen Tu 0c13d7e472 travis.yml: don't die when diff reports diff 2015-02-14 22:46:56 -08:00
Stephen Tu 6ef2e47cd8 travis.yml: add custom config.hpp 2015-02-14 22:43:46 -08:00
Stephen Tu d346878ef9 travis.yml: we really want config.hpp 2015-02-14 22:35:51 -08:00
Stephen Tu 0c6aa64468 travis.yml: cat arma_config.hpp for debug 2015-02-14 22:33:09 -08:00
Stephen Tu 3b9cfef3e0 travis.yml: hopefully this includes all the relevant boost libs 2015-02-14 22:29:33 -08:00
Stephen Tu 76ca1cf203 travis.yml: add a newer version of boost 2015-02-14 22:27:38 -08:00
Stephen Tu 8d22eb5da0 travis.yml: add libboost-dev dep 2015-02-14 22:24:57 -08:00
Stephen Tu f34f516b8a travis.yml: maybe this will let us get the right version of arma? 2015-02-14 22:21:39 -08:00
Stephen Tu c04b6af570 travis.yml: add armadillo dep 2015-02-14 22:17:49 -08:00