SvmModel Class Reference

Abstract class for different SVM models. More...

#include <svmmodel.h>

Inheritance diagram for SvmModel:

Public Member Functions
virtual void	train (const DataSet &dataset, SvmParam param)=0
virtual vector< float_type >	predict (const DataSet::node2d &instances, int batch_size)
void	predict_dec_values (const DataSet::node2d &instances, SyncArray< float_type > &dec_values, int batch_size) const
virtual vector< float_type >	cross_validation (DataSet dataset, SvmParam param, int n_fold)
virtual void	save_to_file (string path)
virtual void	load_from_file (string path)

Protected Member Functions
virtual void	model_setup (const DataSet &dataset, SvmParam &param)

Protected Attributes
SvmParam	param
SyncArray< float_type >	coef
DataSet::node2d	sv
SyncArray< int >	n_sv
	the number of support vectors for each class
int	n_total_sv
	the number of support vectors for all classes
SyncArray< float_type >	rho
	the bias term for each binary model
int	n_classes = 2
	the number of classes
size_t	n_binary_models
	the number of binary models, equal to \(k(k-1)/2\), where \(k\) is the number of classes
vector< float_type >	probA
	be used to predict probability for each binary model
vector< float_type >	probB
	be used to predict probability for each binary model
vector< int >	label
	only for SVC, maps logical label (0,1,2,...) to real label in dataset (maybe 2,4,5,...)

Detailed Description

Abstract class for different SVM models.

Member Function Documentation

cross_validation()

vector< float_type > SvmModel::cross_validation ( DataSet  dataset, SvmParam  param, int  n_fold  )

virtual

performing cross-validation. In \(k\)-fold cross_validation, dataset is spilt into \(k\) equal size parts. Then each part is used as testing set, while the remaining \(k-1\) parts are used as training set. The whole dataset will be predicted after \(k\) training and testing.

Parameters

dataset	training dataset
param	param for cross-validation
n_fold	the number of fold in cross-validation

Returns

the same structure as predict()

load_from_file()

void SvmModel::load_from_file ( string  path )

virtual

load SvmModel from a file, the file format is the same as LIBSVM. before call this function, one must read the first line of the model file, which contains svm_type. Then construct right SvmModel (SVC, SVR, ...), since SvmModel is an abstract class.

Parameters

path	path and filename of saved model file

model_setup()

void SvmModel::model_setup ( const DataSet &  dataset, SvmParam &  param  )

protectedvirtual

called at the begining of train(), do initialization

Parameters

dataset
param

Reimplemented in SVC.

predict()

vector< float_type > SvmModel::predict ( const DataSet::node2d &  instances, int  batch_size  )

virtual

predict label given instances.

Parameters

instances	instances used
batch_size	the number of instances to predict parallel, higher value needs more memory

Returns

label (SVC, NuSVC), real number (SVR, NuSVR), {-1,+1} (OneClassSVC)

Reimplemented in SVC, and OneClassSVC.

predict_dec_values()

void SvmModel::predict_dec_values	(	const DataSet::node2d &	instances,
		SyncArray< float_type > &	dec_values,
		int	batch_size
	)		const

predict decision values.

Parameters

[in]	instances	instances used
[out]	dec_values	decision values predicted, #instances \(times\) n_binary_models array
[in]	batch_size	the number of instances to predict parallel, higher value needs more memory

save_to_file()

void SvmModel::save_to_file ( string  path )

virtual

save SvmModel to a file, the file format is the same as LIBSVM

Parameters

path	path and filename of the model file to save

train()

virtual void SvmModel::train ( const DataSet &  dataset, SvmParam  param  )

pure virtual

train model given dataset and param.

Parameters

dataset	training dataset
param	param for training

Implemented in SVC, SVR, NuSVR, and OneClassSVC.

Member Data Documentation

coef

SyncArray<float_type> SvmModel::coef

protected

coefficients for each support vector, the structure is the same as LIBSVM. The coefficient is equal to \(\alpha_iy_i\) in SVM dual optimization problem. For one-vs-one multi-class decomposition, \(k\) is the number of classes, and \(N_{sv}\) is the number of support vectors. The size of coef is \( (k-1)\times N_{sv}\). Each support vector has at most \(k-1\) coefficients. For each binary classifier \(SVM_{i,j}\), the coefficients locate in: (1) class i, coef[j-1][sv_start[i]...]; (2) class j, coef[i][sv_start[j]...], where sv_start[i] is the start position of support vectors of class i.

sv

DataSet::node2d SvmModel::sv

protected

support vectors of this model. The support vectors is grouped in classes 0,1,2,.... The sequence of them in each group is as they appear in original dataset. A training instance is saved as a support vector IFF it is a support vector in at least one binary model.

---

The documentation for this class was generated from the following files:

• include/thundersvm/model/svmmodel.h
• src/thundersvm/model/svmmodel.cpp