LightGBMWrapper - FIX some variables names and types and codestyle

Author: hudlijac

cmake/dependencies.cmake

@@ -2,6 +2,7 @@
 find_package(Armadillo REQUIRED)
 find_package(Boost REQUIRED COMPONENTS regex serialization)
 find_package(LIGHTGBM REQUIRED)
+find_package(MLPACK REQUIRED)
 find_package(OpenMP REQUIRED)
 find_package(Python3 REQUIRED COMPONENTS Development NumPy)
 

cmake/modules/FindMLPACK.cmake

@@ -0,0 +1,40 @@
+# Find the mlpack includes and library
+#
+# This module defines the following IMPORTED targets:
+#
+#  mlpack::mlpack          - The "mlpack" library, if found.
+#
+# This module will set the following variables in your project:
+#
+#  MLPACK_INCLUDE_DIRS - where to find <mlpack/core.hpp>, etc.
+#  MLPACK_FOUND        - True if the mlpack library has been found.
+
+# Use pkg-config (if available) to get the library directories and then use
+# these values as hints for find_path() and find_library() functions.
+find_package(PkgConfig QUIET)
+if (PKG_CONFIG_FOUND)
+	pkg_check_modules(PC_MLPACK QUIET mlpack)
+endif()
+
+find_path(
+	MLPACK_INCLUDE_DIR mlpack
+	HINTS ${PC_MLPACK_INCLUDEDIR} ${PC_MLPACK_INCLUDE_DIRS}
+	PATH_SUFFIXES include
+)
+
+if (PC_MLPACK_VERSION)
+	# Version extracted from pkg-config
+	set(MLPACK_VERSION_STRING ${PC_MLPACK_VERSION})
+endif()
+
+# Handle find_package() arguments (i.e. QUIETLY and REQUIRED) and set
+# MLPACK_FOUND to TRUE if all listed variables are filled.
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(
+	MLPACK
+	REQUIRED_VARS MLPACK_INCLUDE_DIR
+	VERSION_VAR MLPACK_VERSION_STRING
+)
+
+set(MLPACK_INCLUDE_DIRS ${MLPACK_INCLUDE_DIR})
+mark_as_advanced(MLPACK_INCLUDE_DIR)

include/wif/ml/lightGBMWrapper.hpp

@@ -99,21 +99,18 @@ class LightGBMWrapper {
 	 * @brief Train LightGBM model
 	 *
 	 * @param datasetFileName is the name of the file with training data
-	 * @param datasetParams are the additional parameters of the training dataset. Default:
-	 * eader=true label=name:label
+	 * @param datasetParams are the additional parameters of the training dataset
 	 * @param numOfIterations is the number of training iterations (how many
 	 * LGBM_BoosterUpdateOneIter function is called, see
-	 * https://lightgbm.readthedocs.io/en/stable/C-API.html). Default: 100
+	 * https://lightgbm.readthedocs.io/en/stable/C-API.html)
 	 * @param params are parameters in format ‘key1=value1 key2=value2’ (see
-	 * https://lightgbm.readthedocs.io/en/stable/C-API.html). Default: boosting_type=gbdt
-	 * objective=binary metric=auc num_leaves=30 learning_rate=0.05 feature_fraction=0.9
-	 * @param modelFileName name of the file where the trained model will be saved. Default:
-	 * model.txt
+	 * https://lightgbm.readthedocs.io/en/stable/C-API.html)
+	 * @param modelFileName name of the file where the trained model will be saved
 	 */
 	void train(
-		const std::string datasetFileName,
+		const std::string& datasetFileName,
 		const char* datasetParams = "header=true label=name:label",
-		const int numOfIterations = 100,
+		const unsigned numOfIterations = 100,
 		const char* params
 		= "boosting_type=gbdt objective=binary metric=auc num_leaves=30 learning_rate=0.05 "
 		  "feature_fraction=0.9",

src/wif/ml/lightGBMWrapper.cpp

@@ -42,7 +42,6 @@ bool LightGBMWrapper::loadModel(const std::string& modelPath)
 	if (!LGBM_BoosterCreateFromModelfile(modelPath.c_str(), &m_outNumIterations, &m_booster)) {
 		m_isLoaded = true;
 		m_modelPath = modelPath;
-
 		return true;
 	}
 
@@ -54,14 +53,11 @@ ClfResult LightGBMWrapper::classify(const FlowFeatures& flowFeatures)
 	std::vector<double> dataToClassify; // classified features from flowfeatures are extracted here
 	int64_t outLen; // length of output result
 	int numOfClasses; // number of classes
-
 	LGBM_BoosterGetNumClasses(m_booster, &numOfClasses);
 
 	std::vector<double> pred(numOfClasses); // vector with predictions
-
 	for (const auto& featureID : m_featureIDs) {
 		double value = flowFeatures.get<double>(featureID);
-
 		dataToClassify.push_back(value);
 	}
 
@@ -86,8 +82,7 @@ std::vector<ClfResult> LightGBMWrapper::classify(const std::vector<FlowFeatures>
 	std::vector<double> dataToClassify; // Classified features from flowfeatures are extracted here
 	int64_t outLen; // length of output result
 	int numOfClasses; // number of classes
-	std::vector<ClfResult>
-		burstResults; // vector with predictions in ClfResult format for return value
+	std::vector<ClfResult> burstResults; // vector with predictions in ClfResult format
 
 	burstResults.reserve(burstOfFeatures.size());
 	LGBM_BoosterGetNumClasses(m_booster, &numOfClasses);
@@ -97,7 +92,6 @@ std::vector<ClfResult> LightGBMWrapper::classify(const std::vector<FlowFeatures>
 	for (const auto& feature : burstOfFeatures) { // data preparation for classification
 		for (const auto& featureId : m_featureIDs) {
 			double value = feature.get<double>(featureId);
-
 			dataToClassify.push_back(value);
 		}
 	}
@@ -116,10 +110,10 @@ std::vector<ClfResult> LightGBMWrapper::classify(const std::vector<FlowFeatures>
 		&outLen,
 		pred.data());
 
-	for (size_t i = 0; i < burstOfFeatures.size(); ++i) { // converting pred to burstResults
+	for (unsigned idx = 0; idx < burstOfFeatures.size(); ++idx) { // converting pred to burstResults
 		std::vector<double> probabilities(
-			pred.begin() + i * numOfClasses,
-			pred.begin() + (i + 1) * numOfClasses);
+			pred.begin() + idx * numOfClasses,
+			pred.begin() + (idx + 1) * numOfClasses);
 		burstResults.emplace_back(probabilities);
 	}
 
@@ -132,9 +126,9 @@ bool LightGBMWrapper::isLoaded() const
 }
 
 void LightGBMWrapper::train(
-	const std::string datasetFileName,
+	const std::string& datasetFileName,
 	const char* datasetParams,
-	const int numOfIterations,
+	const unsigned numOfIterations,
 	const char* params,
 	const std::string modelFileName)
 {
@@ -152,7 +146,7 @@ void LightGBMWrapper::train(
 		throw std::runtime_error("Error creating booster");
 	}
 
-	for (int i = 0; i < numOfIterations; ++i) { // training
+	for (unsigned i = 0; i < numOfIterations; ++i) { // training
 		int isFinished;
 		LGBM_BoosterUpdateOneIter(m_booster, &isFinished);
 		if (isFinished) {