Support vector machine model classifier. More...

#include <config.h>
#include <stdlib.h>
#include <stdio.h>
#include <inttypes.h>
#include <assert.h>
#include <string.h>
#include <errno.h>
#include <math.h>
#include <libxml/tree.h>
#include <libxml/parser.h>
#include <libxml/valid.h>
#include <jwsc/base/error.h>
#include <jwsc/base/limits.h>
#include <jwsc/base/file_io.h>
#include <jwsc/vector/vector.h>
#include <jwsc/vector/vector_io.h>
#include <jwsc/vector/vector_math.h>
#include <jwsc/matrix/matrix.h>
#include <jwsc/matrix/matrix_io.h>
#include <jwsc/matblock/matblock.h>
#include <jwsc/matblock/matblock_io.h>
#include <jwsc/stat/gmm.h>
#include "xml.h"
#include "haplo_groups.h"
#include "svm_tree.h"

Functions
void	train_svm_model (SVM_model *model_out, const Vector_u32 labels, const Matrix_i32 *markers, double cost, double gamma)
	Trains a SVM model.
Error *	predict_label_with_svm_model (uint32_t label_out, double confidence_out, const struct svm_node markers, const SVM_model *model)
	Predicts the label for a marker sample using a SVM model.
static void	get_predicted_labels_confidence (Vector_d *confidence_out, const Vector_u32 labels_v, const Matrix_d confidence_matrix, const SVM_model model)
	Gets the confidence value for each predicted label.
Error *	predict_labels_with_svm_model (Vector_u32 labels_out, Vector_d confidence_out, const Matrix_i32 markers, const SVM_model model)
	Predicts the labels for a set of marker samples using a SVM model.
Error *	read_svm_model (SVM_model *model_out, const char fname)
	Reads a SVM model.
Error *	write_svm_model (SVM_model model, const char fname)
	Writes a SVM model.
Error *	write_svm_model_training_data (const Vector_u32 labels, const Matrix_i32 markers, const char *fname)
	Writes the training data for an SVM model to a file.
void	free_svm_model (SVM_model *model)
	Frees a SVM model.
static Error *	read_svm_xml_doc (xmlDoc *xml_doc_out, const char xml_fname, const char *dtd_fname)
	Reads and and optionally validates an XML document.
static Error *	create_svm_model_tree_from_xml_node (SVM_model_tree *tree_out, const SVM_model_tree parent, uint32_t parent_label, xmlNode *xml_node)
	Allocates and initializes a model tree from an xml node.
static Error *	create_svm_model_node_from_xml_node (xmlNode xml_node, SVM_model_node svm_node, uint32_t i)
	Fills in an SVM node from an xml node.
static Error *	create_model_training_data (Vector_u32 train_labels_out, Matrix_i32 train_markers_out, const Vector_u32 data_labels, const Matrix_i32 data_markers, uint32_t model_labels, const Vector_u32 *model_altlabels)
	Creates a set of training data as a model-specific labeled subset of a larger data set.
static Error *	train_svm_model_node (SVM_model_node node, const Vector_u32 labels, const Matrix_i32 *markers)
	Trains a model in the tree using node-specific data.
Error *	train_svm_model_tree (SVM_model_tree *tree_out, const Vector_u32 labels, const Matrix_i32 markers, const char tree_xml_fname, const char *tree_dtd_fname)
	Trains a SVM model tree .
static double	get_predicted_label_confidence (uint32_t label, double confs, const SVM_model model)
	Gets the confidence value for a predicted label.
static uint32_t	get_best_one_against_all_model (const SVM_model_tree tree, Vector_u32 labels, Vector_d *confs)
	Gets the best SVM model (index).
static uint32_t	get_best_one_against_one_model (const SVM_model_tree tree, Vector_u32 labels, Vector_d *confs)
	Gets the best SVM model (index).
static Error *	recursively_predict_label_in_model_tree (uint32_t label_out, double confidence_out, const SVM_model_tree tree, const struct svm_node markers)
	Recursively predicts a label from a model tree.
Error *	predict_labels_with_svm_model_tree (Vector_u32 labels_out, Vector_d confidence_out, const Matrix_i32 markers, const SVM_model_tree tree)
	Predicts the labels for a set of marker samples using a SVM model tree.
static Error *	read_svm_model_node (SVM_model_node node, const char model_dirname)
	Recursively reads the model in a node from its file name.
Error *	read_svm_model_tree (SVM_model_tree *tree_out, const char tree_xml_fname, const char tree_dtd_fname, const char model_dirname)
	Reads a SVM model tree from.
Error *	write_svm_model_tree (const SVM_model_tree tree, const char model_dirname)
	Writes a SVM model tree.
static Error *	write_svm_model_node_training_data (SVM_model_node node, const Vector_u32 labels, const Matrix_i32 markers, const char data_dirname)
	Writes the training data for a model in the tree using node-specific data.
Error *	write_svm_model_tree_training_data (const Vector_u32 labels, const Matrix_i32 markers, const char tree_xml_fname, const char tree_dtd_fname, const char *data_dirname)
	Writes the training data for an SVM model tree to a file.
void	free_svm_model_tree (SVM_model_tree *tree)
	Frees a SVM model tree.

Detailed Description

Support vector machine model classifier.

Author:: Joseph Schlecht

License:: Creative Commons BY-NC-SA 3.0

Definition in file svm_tree.c.

Function Documentation

void train_svm_model	(	SVM_model **	model_out,
		const Vector_u32 *	labels,
		const Matrix_i32 *	markers,
		double	cost,
		double	gamma
	)

Trains a SVM model.

Parameters:

model_out	Result parameter. If *model_out is NULL, a model is allocated; otherwise its space is re-used.
labels	Labels for training, with ith element as corresponding to the ith sample in markers.
markers	Markers for training, with ith row as a sample corresponding to the ith label in labels.
cost	Cost parameter for the SVM model.
gamma	Gamma parameter for the SVM model.

Definition at line 98 of file svm_tree.c.

Error* predict_label_with_svm_model	(	uint32_t *	label_out,
		double **	confidence_out,
		const struct svm_node *	markers,
		const SVM_model *	model
	)

Predicts the label for a marker sample using a SVM model.

Parameters:

label_out	Result parameter.
confidence_out	Result parameter. Must be an array of doubles of length svm_get_nr_class(model).
markers	Marker data to predict.
model	Trained model to use for predicting.

Returns:: On success, NULL is returned; otherwise an error is returned.

Definition at line 172 of file svm_tree.c.

static void get_predicted_labels_confidence	(	Vector_d **	confidence_out,
		const Vector_u32 *	labels_v,
		const Matrix_d *	confidence_matrix,
		const SVM_model *	model
	)		`[static]`

Gets the confidence value for each predicted label.

Definition at line 188 of file svm_tree.c.

Error* predict_labels_with_svm_model	(	Vector_u32 **	labels_out,
		Vector_d **	confidence_out,
		const Matrix_i32 *	markers,
		const SVM_model *	model
	)

Predicts the labels for a set of marker samples using a SVM model.

Parameters:

labels_out	Result parameter. If *labels_out is NULL, it is allocated; otherwise its space is re-used.
confidence_out	Result parameter. If *confidence_out is NULL, it is allocated; otherwise its space is re-used.
markers	Marker data to predict. Each row is a sample for prediction, corresponding to an element in the result parameters.
model	Trained model to use for predicting.

Returns:: On success, NULL is returned; otherwise an error is returned, but the result parameters are not freed.

Definition at line 241 of file svm_tree.c.

Error* read_svm_model	(	SVM_model **	model_out,
		const char *	fname
	)

Reads a SVM model.

Parameters:

model_out	Result parameter. If *model_out is NULL, a model is allocated; otherwise its space is re-used.
fname	File to read the model from.

Returns:: On success, NULL is returned; otherwise an error is returned but does not free result parameter.

Definition at line 299 of file svm_tree.c.

Error* write_svm_model	(	SVM_model *	model,
		const char *	fname
	)

Writes a SVM model.

Parameters:

model	Model to write.
fname	File to write to.

Returns:: On success, NULL is returned; otherwise an error is returned.

Definition at line 328 of file svm_tree.c.

Error* write_svm_model_training_data	(	const Vector_u32 *	labels,
		const Matrix_i32 *	markers,
		const char *	fname
	)

Writes the training data for an SVM model to a file.

The format is txt-based and consistent with input to libsvm.

  label-index  1:marker-val 2:marker-val 3:marker-val ...

Parameters:

labels	Model training labels to write.
markers	Model training markers to write.
fname	File to write to.

Returns:: On success, NULL is returned; otherwise an error is returned.

Definition at line 357 of file svm_tree.c.

void free_svm_model ( SVM_model * model )

Frees a SVM model.

Parameters:

model Model to free.

Definition at line 399 of file svm_tree.c.

static Error* read_svm_xml_doc	(	xmlDoc **	xml_doc_out,
		const char *	xml_fname,
		const char *	dtd_fname
	)		`[static]`

Reads and and optionally validates an XML document.

Definition at line 425 of file svm_tree.c.

static Error * create_svm_model_tree_from_xml_node	(	SVM_model_tree **	tree_out,
		const SVM_model_tree *	parent,
		uint32_t	parent_label,
		xmlNode *	xml_node
	)		`[static]`

Allocates and initializes a model tree from an xml node.

Does not train or read the model; leaves the model field as NULL in each tree node.

Definition at line 600 of file svm_tree.c.

static Error* create_svm_model_node_from_xml_node	(	xmlNode *	xml_node,
		SVM_model_node *	svm_node,
		uint32_t	i
	)		`[static]`

Fills in an SVM node from an xml node.

Definition at line 480 of file svm_tree.c.

static Error* create_model_training_data	(	Vector_u32 **	train_labels_out,
		Matrix_i32 **	train_markers_out,
		const Vector_u32 *	data_labels,
		const Matrix_i32 *	data_markers,
		uint32_t *	model_labels,
		const Vector_u32 **	model_altlabels
	)		`[static]`

Creates a set of training data as a model-specific labeled subset of a larger data set.

Definition at line 687 of file svm_tree.c.

static Error* train_svm_model_node	(	SVM_model_node *	node,
		const Vector_u32 *	labels,
		const Matrix_i32 *	markers
	)		`[static]`

Trains a model in the tree using node-specific data.

Definition at line 768 of file svm_tree.c.

Error* train_svm_model_tree	(	SVM_model_tree **	tree_out,
		const Vector_u32 *	labels,
		const Matrix_i32 *	markers,
		const char *	tree_xml_fname,
		const char *	tree_dtd_fname
	)

Trains a SVM model tree .

Parameters:

tree_out	Result parameter.
labels	Sample group labels.
markers	Sample marker values.
tree_xml_fname	XML file containing the model tree information.
tree_dtd_fname	DTD file for validating the XML file, can be NULL.

Definition at line 829 of file svm_tree.c.

static double get_predicted_label_confidence	(	uint32_t	label,
		double *	confs,
		const SVM_model *	model
	)		`[static]`

Gets the confidence value for a predicted label.

Definition at line 884 of file svm_tree.c.

static uint32_t get_best_one_against_all_model	(	const SVM_model_tree *	tree,
		Vector_u32 *	labels,
		Vector_d *	confs
	)		`[static]`

Gets the best SVM model (index).

Definition at line 914 of file svm_tree.c.

static uint32_t get_best_one_against_one_model	(	const SVM_model_tree *	tree,
		Vector_u32 *	labels,
		Vector_d *	confs
	)		`[static]`

Gets the best SVM model (index).

Definition at line 1015 of file svm_tree.c.

static Error* recursively_predict_label_in_model_tree	(	uint32_t *	label_out,
		double *	confidence_out,
		const SVM_model_tree *	tree,
		const struct svm_node *	markers
	)		`[static]`

Recursively predicts a label from a model tree.

Definition at line 1083 of file svm_tree.c.

Error* predict_labels_with_svm_model_tree	(	Vector_u32 **	labels_out,
		Vector_d **	confidence_out,
		const Matrix_i32 *	markers,
		const SVM_model_tree *	tree
	)

Predicts the labels for a set of marker samples using a SVM model tree.

Parameters:

labels_out	Result parameter. If *labels_out is NULL, it is allocated; otherwise its space is re-used.
confidence_out	Result parameter. If *confidence_out is NULL, it is allocated; otherwise its space is re-used.
markers	Marker data to predict. Each row is a sample for prediction, corresponding to an element in the result parameters.
tree	Trained model tree to use for predicting.

Returns:: On success, NULL is returned; otherwise an error is returned, but the result parameters are not freed.

Definition at line 1170 of file svm_tree.c.

static Error* read_svm_model_node	(	SVM_model_node *	node,
		const char *	model_dirname
	)		`[static]`

Recursively reads the model in a node from its file name.

Definition at line 1223 of file svm_tree.c.

Error* read_svm_model_tree	(	SVM_model_tree **	tree_out,
		const char *	tree_xml_fname,
		const char *	tree_dtd_fname,
		const char *	model_dirname
	)

Reads a SVM model tree from.

Parameters:

tree_out	Result parameter.
tree_xml_fname	XML file containing the model tree information.
tree_dtd_fname	DTD file for validating the XML file, can be NULL.
model_dirname	Directory prefixing each model file in the tree.

Definition at line 1264 of file svm_tree.c.

Error* write_svm_model_tree	(	const SVM_model_tree *	tree,
		const char *	model_dirname
	)

Writes a SVM model tree.

Parameters:

tree	Model tree to write.
model_dirname	Directory to prefix to each model file in the tree.

Definition at line 1321 of file svm_tree.c.

static Error* write_svm_model_node_training_data	(	SVM_model_node *	node,
		const Vector_u32 *	labels,
		const Matrix_i32 *	markers,
		const char *	data_dirname
	)		`[static]`

Writes the training data for a model in the tree using node-specific data.

Definition at line 1364 of file svm_tree.c.

Error* write_svm_model_tree_training_data	(	const Vector_u32 *	labels,
		const Matrix_i32 *	markers,
		const char *	tree_xml_fname,
		const char *	tree_dtd_fname,
		const char *	data_dirname
	)

Writes the training data for an SVM model tree to a file.

Writes the data files using the name of the model to the data_dir directory.

Parameters:

labels	Sample group labels.
markers	Sample marker values.
tree_xml_fname	XML file containing the model tree information.
tree_dtd_fname	DTD file for validating the XML file, can be NULL.
data_dirname	Directory to write the training data files.

Definition at line 1436 of file svm_tree.c.

void free_svm_model_tree ( SVM_model_tree * tree )

Frees a SVM model tree.

Parameters:

tree	Model tree to free.

Definition at line 1493 of file svm_tree.c.

Functions

Detailed Description

Function Documentation