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

removed extra L-BFGS related parameters

Browse Source
This commit is contained in:
Manish
2018-06-28 15:45:40 +05:30
parent bc3b5bba37
commit fddba649b1
3 changed files with 6 additions and 122 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/mlpack/methods/lmnn/lmnn_main.cpp
+4 -29
View File
@@ -94,12 +94,9 @@ PROGRAM_INFO("Large Margin Nearest Neighbors (LMNN)",
"The L-BFGS optimizer, specified by the value 'lbfgs' for the parameter " +
PRINT_PARAM_STRING("optimizer") + ", uses a back-tracking line search "
"algorithm to minimize a function. The following parameters are used by "
"L-BFGS: " + PRINT_PARAM_STRING("num_basis") + " (specifies the number"
" of memory points used by L-BFGS), " +
PRINT_PARAM_STRING("max_iterations") + ", " +
PRINT_PARAM_STRING("armijo_constant") + ", " +
PRINT_PARAM_STRING("wolfe") + ", " + PRINT_PARAM_STRING("tolerance") +
" (the optimization is terminated when the gradient norm is below this "
"L-BFGS: " + PRINT_PARAM_STRING("max_iterations") + ", " +
PRINT_PARAM_STRING("tolerance") +
"(the optimization is terminated when the gradient norm is below this "
"value). For more details on the L-BFGS optimizer, consult either the "
"mlpack L-BFGS documentation (in lbfgs.hpp) or the vast set of published "
"literature on L-BFGS. In addition, a normalized starting point can be "
@@ -135,7 +132,7 @@ PARAM_STRING_IN("optimizer", "Optimizer to use; 'amsgrad', 'bbsgd', 'sgd', or "
"'lbfgs'.", "O", "amsgrad");
PARAM_DOUBLE_IN("regularization", "Regularization for LMNN objective function ",
"r", 0.5);
PARAM_INT_IN("rank", "Rank of distance matrix to be optimized. ", "K", 0);
PARAM_INT_IN("rank", "Rank of distance matrix to be optimized. ", "A", 0);
PARAM_FLAG("normalize", "Use a normalized starting point for optimization. It"
"is useful for when points are far apart, or when SGD is returning NaN.",
"N");
@@ -150,10 +147,6 @@ PARAM_DOUBLE_IN("step_size", "Step size for AMSGrad, BB_SGD and SGD (alpha).",
PARAM_FLAG("linear_scan", "Don't shuffle the order in which data points are "
"visited for SGD or mini-batch SGD.", "L");
PARAM_INT_IN("batch_size", "Batch size for mini-batch SGD.", "b", 50);
PARAM_INT_IN("num_basis", "Number of memory points to be stored for L-BFGS.",
"B", 5);
PARAM_DOUBLE_IN("armijo_constant", "Armijo constant for L-BFGS.", "A", 1e-4);
PARAM_DOUBLE_IN("wolfe", "Wolfe condition parameter for L-BFGS.", "w", 0.9);
PARAM_INT_IN("range", "Number of iterations after which impostors needs to be "
"recalculated", "R", 1);
PARAM_INT_IN("seed", "Random seed. If 0, 'std::time(NULL)' is used.", "s", 0);
@@ -224,26 +217,14 @@ static void mlpackMain()
// Warn on unused parameters.
if (optimizerType == "amsgrad")
{
ReportIgnoredParam("num_basis", "L-BFGS optimizer is not being used");
ReportIgnoredParam("armijo_constant", "L-BFGS optimizer is not being used");
ReportIgnoredParam("wolfe", "L-BFGS optimizer is not being used");
ReportIgnoredParam("num_basis", "L-BFGS optimizer is not being used");
ReportIgnoredParam("max_iterations", "L-BFGS optimizer is not being used");
}
else if (optimizerType == "bbsgd")
{
ReportIgnoredParam("num_basis", "L-BFGS optimizer is not being used");
ReportIgnoredParam("armijo_constant", "L-BFGS optimizer is not being used");
ReportIgnoredParam("wolfe", "L-BFGS optimizer is not being used");
ReportIgnoredParam("num_basis", "L-BFGS optimizer is not being used");
ReportIgnoredParam("max_iterations", "L-BFGS optimizer is not being used");
}
else if (optimizerType == "sgd")
{
ReportIgnoredParam("num_basis", "L-BFGS optimizer is not being used");
ReportIgnoredParam("armijo_constant", "L-BFGS optimizer is not being used");
ReportIgnoredParam("wolfe", "L-BFGS optimizer is not being used");
ReportIgnoredParam("num_basis", "L-BFGS optimizer is not being used");
ReportIgnoredParam("max_iterations", "L-BFGS optimizer is not being used");
}
else if (optimizerType == "lbfgs")
@@ -283,9 +264,6 @@ static void mlpackMain()
const bool normalize = CLI::HasParam("normalize");
const bool printAccuracy = CLI::HasParam("print_accuracy");
const bool shuffle = !CLI::HasParam("linear_scan");
const int numBasis = CLI::GetParam<int>("num_basis");
const double armijoConstant = CLI::GetParam<double>("armijo_constant");
const double wolfe = CLI::GetParam<double>("wolfe");
const size_t batchSize = (size_t) CLI::GetParam<int>("batch_size");
const size_t range = (size_t) CLI::GetParam<int>("range");
const size_t rank = (size_t) CLI::GetParam<int>("rank");
@@ -388,10 +366,7 @@ static void mlpackMain()
LMNN<LMetric<2>, L_BFGS> lmnn(data, labels, k);
lmnn.Regularization() = regularization;
lmnn.Range() = range;
lmnn.Optimizer().NumBasis() = numBasis;
lmnn.Optimizer().MaxIterations() = maxIterations;
lmnn.Optimizer().ArmijoConstant() = armijoConstant;
lmnn.Optimizer().Wolfe() = wolfe;
lmnn.Optimizer().MinGradientNorm() = tolerance;
lmnn.LearnDistance(distance);
src/mlpack/tests/lmnn_test.cpp
+2 -1
View File
@@ -1,5 +1,6 @@
/**
* @file lmnn_test.cpp
* @author Marcus Edel
* @author Ryan Curtin
* @author Manish Kumar
*
@@ -490,7 +491,7 @@ BOOST_AUTO_TEST_CASE(LMNNLowRankAccuracyTest)
// BOOST_REQUIRE_CLOSE(Accuracy1, Accuracy2, 2.0);
}
// Comprehensive gradient tests by Ryan Curtin.
// Comprehensive gradient tests by Marcus Edel & Ryan Curtin.
// Simple numerical gradient checker.
template<class FunctionType>
src/mlpack/tests/main_tests/lmnn_test.cpp
-92
View File
@@ -608,98 +608,6 @@ BOOST_AUTO_TEST_CASE(LMNNDiffRangeTest)
transformedData), 0);
}
/**
* Ensure that different value of wolfe results in a
* different output matrix.
*/
BOOST_AUTO_TEST_CASE(LMNNDiffWolfeTest)
{
arma::mat inputData;
if (!data::Load("iris.csv", inputData))
BOOST_FAIL("Cannot load iris.csv!");
arma::Row<size_t> labels;
if (!data::Load("iris_labels.txt", labels))
BOOST_FAIL("Cannot load iris_labels.txt!");
// Set parameters.
SetInputParam("input", inputData);
SetInputParam("labels", labels);
SetInputParam("optimizer", std::string("lbfgs"));
SetInputParam("k", 5);
SetInputParam("wolfe", 0.1);
mlpackMain();
arma::mat output = CLI::GetParam<arma::mat>("output");
arma::mat transformedData = CLI::GetParam<arma::mat>("transformed_data");
// Reset settings.
CLI::ClearSettings();
CLI::RestoreSettings(testName);
// Set different parameters.
SetInputParam("input", std::move(inputData));
SetInputParam("labels", std::move(labels));
SetInputParam("optimizer", std::string("lbfgs"));
SetInputParam("k", 5);
SetInputParam("wolfe", 1.0);
mlpackMain();
// Check that the output matrices are different.
BOOST_REQUIRE_GT(
arma::accu(CLI::GetParam<arma::mat>("output") != output), 0);
BOOST_REQUIRE_GT(arma::accu(CLI::GetParam<arma::mat>("transformed_data") !=
transformedData), 0);
}
/**
* Ensure that different value of NumBasis results in a
* different output matrix.
*/
BOOST_AUTO_TEST_CASE(LMNNDiffNumBasisTest)
{
arma::mat inputData;
if (!data::Load("iris.csv", inputData))
BOOST_FAIL("Cannot load iris.csv!");
arma::Row<size_t> labels;
if (!data::Load("iris_labels.txt", labels))
BOOST_FAIL("Cannot load iris_labels.txt!");
// Set parameters and use a larger num_basis.
SetInputParam("input", inputData);
SetInputParam("labels", labels);
SetInputParam("optimizer", std::string("lbfgs"));
SetInputParam("k", 5);
SetInputParam("num_basis", (int) 100);
mlpackMain();
arma::mat output = CLI::GetParam<arma::mat>("output");
arma::mat transformedData = CLI::GetParam<arma::mat>("transformed_data");
// Reset Settings.
CLI::ClearSettings();
CLI::RestoreSettings(testName);
// Set parameters with a smaller num_basis.
SetInputParam("input", std::move(inputData));
SetInputParam("labels", labels);
SetInputParam("optimizer", std::string("lbfgs"));
SetInputParam("k", 5);
SetInputParam("num_basis", (int) 1);
mlpackMain();
// Check that the output matrices are different.
BOOST_REQUIRE_GT(
arma::accu(CLI::GetParam<arma::mat>("output") != output), 0);
BOOST_REQUIRE_GT(arma::accu(CLI::GetParam<arma::mat>("transformed_data") !=
transformedData), 0);
}
/**
* Ensure that using a different value of max_iteration
* results in a different output matrix.