OpenCLTemplate.MachineLearning.SVM Class Reference

This class stores the variables of a SMO problem solution. More...

List of all members.

Classes
class	CLSVMSrc
class	ProblemSolver
	This class aggregates all methods for solving SMO problems.
Public Member Functions
	SVM ()
	Constructor that initializes with 0s.
void	initializeWithZeros ()
	Set all values of the solution to 0.
void	PreCalibrateCfg (float tolPositive, float tolNegative)
	Attempts to pre-calibrate configuration parameters. Finds an alpha that enhances similarities between positive examples and reduces similarities between positive and negative examples. Assumes that decreasing lambda increases kernel match.
float	TrainWithCrossValidation ()
	Trains current SVM with cross-validation, adjusting kernel parameter lambda and box parameter C.
float	TrainWithCrossValidation (float CrossValidationSetPercent, float[] lambdaSet, float[] CSet)
	Trains current SVM with cross-validation, adjusting kernel parameter lambda and box parameter C. Returns best performance so far.
void	Train ()
	Trains current SVM.
void	ResetSolution ()
	Resets current SVM solution.
void	RemoveNonSupportVectors ()
	Removes all vectors that are not SVMs, ie, so that alpha[i]=0.
bool	Classify (TrainingUnit Sample)
	Classifies a training unit as positive or negative (true or false).
float	ClassificationValue (TrainingUnit Sample)
	Classifies a training unit with a float. The bigger, the more positive the sample. Values greater than zero are assumed to be positive samples.
float	GetTrainingSetHitRate ()
	Gets the percentage of training examples classified correctly in the training set.
float	GetTrainingSetHitRate (int ind)
	Gets the percentage of training examples classified correctly starting from the ind-th training sample.
float	GetHitRate (TrainingSet samples)
	Computes hit rates for a given test set.
void	Load (string FileName)
	Copy all values from another solution.
void	Save (string FileName)
	Writes current solution to a file.
	SVM ()
	Constructor that initializes with 0s.
void	initializeWithZeros ()
	Set all values of the solution to 0.
void	PreCalibrateCfg (float tolPositive, float tolNegative)
	Attempts to pre-calibrate configuration parameters. Finds an alpha that enhances similarities between positive examples and reduces similarities between positive and negative examples. Assumes that decreasing lambda increases kernel match.
float	TrainWithCrossValidation ()
	Trains current SVM with cross-validation, adjusting kernel parameter lambda and box parameter C.
float	TrainWithCrossValidation (float CrossValidationSetPercent, float[] lambdaSet, float[] CSet)
	Trains current SVM with cross-validation, adjusting kernel parameter lambda and box parameter C. Returns best performance so far.
void	Train ()
	Trains current SVM.
void	ResetSolution ()
	Resets current SVM solution.
void	RemoveNonSupportVectors ()
	Removes all vectors that are not SVMs, ie, so that alpha[i]=0.
bool	Classify (TrainingUnit Sample)
	Classifies a training unit as positive or negative (true or false).
float	ClassificationValue (TrainingUnit Sample)
	Classifies a training unit with a float. The bigger, the more positive the sample. Values greater than zero are assumed to be positive samples.
float	GetTrainingSetHitRate ()
	Gets the percentage of training examples classified correctly in the training set.
float	GetTrainingSetHitRate (int ind)
	Gets the percentage of training examples classified correctly starting from the ind-th training sample.
float	GetHitRate (TrainingSet samples)
	Computes hit rates for a given test set.
void	Load (string FileName)
	Copy all values from another solution.
void	Save (string FileName)
	Writes current solution to a file.
Static Public Member Functions
static void	CLcalculateAllKernels (SVM problemSolution)
	Computes All kernels and errors accelerating with OpenCL.
static float	CLpredictOutput (SVM problemSolution, TrainingUnit untrainedUnit)
	Predicts the output of a single entry, given a previous problem, solution and correspondent training set.
static float[]	MultiClassify (SVM svm, CLCalc.Program.Image2D Samples)
	Classifies multiple samples stored in OpenCL memory.
static void	CLcalculateAllKernels (SVM problemSolution)
	Computes All kernels and errors accelerating with OpenCL.
static float	CLpredictOutput (SVM problemSolution, TrainingUnit untrainedUnit)
	Predicts the output of a single entry, given a previous problem, solution and correspondent training set.
static float[]	MultiClassify (SVM svm, CLCalc.Program.Image2D Samples)
	Classifies multiple samples stored in OpenCL memory.
Public Attributes
TrainingSet	TrainingSet = null
	Training set for this solution.
ProblemConfig	ProblemCfg = new ProblemConfig(0.1f, 50, 1e-3f, 1)
	Problem configuration.

---

Detailed Description

This class stores the variables of a SMO problem solution.

---

Constructor & Destructor Documentation

OpenCLTemplate.MachineLearning.SVM.SVM

(

)

Constructor that initializes with 0s.

OpenCLTemplate.MachineLearning.SVM.SVM

(

)

Constructor that initializes with 0s.

---

Member Function Documentation

float OpenCLTemplate.MachineLearning.SVM.ClassificationValue

(

TrainingUnit

Sample

)

Classifies a training unit with a float. The bigger, the more positive the sample. Values greater than zero are assumed to be positive samples.

Parameters:

Sample

Sample to be classified

float OpenCLTemplate.MachineLearning.SVM.ClassificationValue

(

TrainingUnit

Sample

)

Classifies a training unit with a float. The bigger, the more positive the sample. Values greater than zero are assumed to be positive samples.

Parameters:

Sample

Sample to be classified

bool OpenCLTemplate.MachineLearning.SVM.Classify

(

TrainingUnit

Sample

)

Classifies a training unit as positive or negative (true or false).

Parameters:

Sample

Sample to be classified

bool OpenCLTemplate.MachineLearning.SVM.Classify

(

TrainingUnit

Sample

)

Classifies a training unit as positive or negative (true or false).

Parameters:

Sample

Sample to be classified

static void OpenCLTemplate.MachineLearning.SVM.CLcalculateAllKernels

(

SVM

problemSolution

)

[static]

Computes All kernels and errors accelerating with OpenCL.

Parameters:

problemSolution

Problem solution SVM

static void OpenCLTemplate.MachineLearning.SVM.CLcalculateAllKernels

(

SVM

problemSolution

)

[static]

Computes All kernels and errors accelerating with OpenCL.

Parameters:

problemSolution

Problem solution SVM

static float OpenCLTemplate.MachineLearning.SVM.CLpredictOutput	(	SVM	problemSolution,
		TrainingUnit	untrainedUnit
	)			[static]

Predicts the output of a single entry, given a previous problem, solution and correspondent training set.

Parameters:

	problemSolution	Correspondent problem solution
	untrainedUnit	Input features from which the output will be predicted

Returns:

The y classification (true/false = positive/negative)

static float OpenCLTemplate.MachineLearning.SVM.CLpredictOutput	(	SVM	problemSolution,
		TrainingUnit	untrainedUnit
	)			[static]

Predicts the output of a single entry, given a previous problem, solution and correspondent training set.

Parameters:

	problemSolution	Correspondent problem solution
	untrainedUnit	Input features from which the output will be predicted

Returns:

The y classification (true/false = positive/negative)

float OpenCLTemplate.MachineLearning.SVM.GetHitRate

(

TrainingSet

samples

)

Computes hit rates for a given test set.

Parameters:

samples

Test set to be used

float OpenCLTemplate.MachineLearning.SVM.GetHitRate

(

TrainingSet

samples

)

Computes hit rates for a given test set.

Parameters:

samples

Test set to be used

float OpenCLTemplate.MachineLearning.SVM.GetTrainingSetHitRate

(

int

ind

)

Gets the percentage of training examples classified correctly starting from the ind-th training sample.

Parameters:

ind

Index to start reading training set samples

float OpenCLTemplate.MachineLearning.SVM.GetTrainingSetHitRate

(

)

Gets the percentage of training examples classified correctly in the training set.

float OpenCLTemplate.MachineLearning.SVM.GetTrainingSetHitRate

(

int

ind

)

Gets the percentage of training examples classified correctly starting from the ind-th training sample.

Parameters:

ind

Index to start reading training set samples

float OpenCLTemplate.MachineLearning.SVM.GetTrainingSetHitRate

(

)

Gets the percentage of training examples classified correctly in the training set.

void OpenCLTemplate.MachineLearning.SVM.initializeWithZeros

(

)

Set all values of the solution to 0.

void OpenCLTemplate.MachineLearning.SVM.initializeWithZeros

(

)

Set all values of the solution to 0.

void OpenCLTemplate.MachineLearning.SVM.Load

(

string

FileName

)

Copy all values from another solution.

Parameters:

FileName

File containing alpha's data

void OpenCLTemplate.MachineLearning.SVM.Load

(

string

FileName

)

Copy all values from another solution.

Parameters:

FileName

File containing alpha's data

static float [] OpenCLTemplate.MachineLearning.SVM.MultiClassify	(	SVM	svm,
		CLCalc.Program.Image2D	Samples
	)			[static]

Classifies multiple samples stored in OpenCL memory.

Parameters:

	Samples	Samples data to classify
	svm	SVM to use as classifier

static float [] OpenCLTemplate.MachineLearning.SVM.MultiClassify	(	SVM	svm,
		CLCalc.Program.Image2D	Samples
	)			[static]

Classifies multiple samples stored in OpenCL memory.

Parameters:

	Samples	Samples data to classify
	svm	SVM to use as classifier

void OpenCLTemplate.MachineLearning.SVM.PreCalibrateCfg	(	float	tolPositive,
		float	tolNegative
	)

Attempts to pre-calibrate configuration parameters. Finds an alpha that enhances similarities between positive examples and reduces similarities between positive and negative examples. Assumes that decreasing lambda increases kernel match.

Parameters:

	tolPositive	Positive kernels average should be greater than tolPositive
	tolNegative	Negative kernels average should be lesser than tolNegative

void OpenCLTemplate.MachineLearning.SVM.PreCalibrateCfg	(	float	tolPositive,
		float	tolNegative
	)

Attempts to pre-calibrate configuration parameters. Finds an alpha that enhances similarities between positive examples and reduces similarities between positive and negative examples. Assumes that decreasing lambda increases kernel match.

Parameters:

	tolPositive	Positive kernels average should be greater than tolPositive
	tolNegative	Negative kernels average should be lesser than tolNegative

void OpenCLTemplate.MachineLearning.SVM.RemoveNonSupportVectors

(

)

Removes all vectors that are not SVMs, ie, so that alpha[i]=0.

void OpenCLTemplate.MachineLearning.SVM.RemoveNonSupportVectors

(

)

Removes all vectors that are not SVMs, ie, so that alpha[i]=0.

void OpenCLTemplate.MachineLearning.SVM.ResetSolution

(

)

Resets current SVM solution.

void OpenCLTemplate.MachineLearning.SVM.ResetSolution

(

)

Resets current SVM solution.

void OpenCLTemplate.MachineLearning.SVM.Save

(

string

FileName

)

Writes current solution to a file.

void OpenCLTemplate.MachineLearning.SVM.Save

(

string

FileName

)

Writes current solution to a file.

void OpenCLTemplate.MachineLearning.SVM.Train

(

)

Trains current SVM.

void OpenCLTemplate.MachineLearning.SVM.Train

(

)

Trains current SVM.

float OpenCLTemplate.MachineLearning.SVM.TrainWithCrossValidation	(	float	CrossValidationSetPercent,
		float[]	lambdaSet,
		float[]	CSet
	)

Trains current SVM with cross-validation, adjusting kernel parameter lambda and box parameter C. Returns best performance so far.

Parameters:

	CrossValidationSetPercent	Percentage of training examples that should be used as cross validation set
	lambdaSet	Values of lambda to try
	CSet	Values of c to try

float OpenCLTemplate.MachineLearning.SVM.TrainWithCrossValidation

(

)

Trains current SVM with cross-validation, adjusting kernel parameter lambda and box parameter C.

float OpenCLTemplate.MachineLearning.SVM.TrainWithCrossValidation	(	float	CrossValidationSetPercent,
		float[]	lambdaSet,
		float[]	CSet
	)

Trains current SVM with cross-validation, adjusting kernel parameter lambda and box parameter C. Returns best performance so far.

Parameters:

	CrossValidationSetPercent	Percentage of training examples that should be used as cross validation set
	lambdaSet	Values of lambda to try
	CSet	Values of c to try

float OpenCLTemplate.MachineLearning.SVM.TrainWithCrossValidation

(

)

Trains current SVM with cross-validation, adjusting kernel parameter lambda and box parameter C.

---

Member Data Documentation

ProblemConfig OpenCLTemplate::MachineLearning.SVM::ProblemCfg = new ProblemConfig(0.1f, 50, 1e-3f, 1)

Problem configuration.

TrainingSet OpenCLTemplate::MachineLearning.SVM::TrainingSet = null

Training set for this solution.