weka.classifiers.functions.supportVector

Class CachedKernel

java.lang.Object

weka.classifiers.functions.supportVector.Kernel

weka.classifiers.functions.supportVector.CachedKernel

All Implemented Interfaces:

Serializable, CapabilitiesHandler, OptionHandler, RevisionHandler

Direct Known Subclasses:

PolyKernel, Puk, RBFKernel

public abstract class CachedKernel
extends Kernel

Base class for RBFKernel and PolyKernel that implements a simple LRU. (least-recently-used) cache if the cache size is set to a value > 0. Otherwise it uses a full cache.

Version:

$Revision: 4549 $

Author:

Eibe Frank (eibe@cs.waikato.ac.nz), Shane Legg (shane@intelligenesis.net) (sparse vector code), Stuart Inglis (stuart@reeltwo.com) (sparse vector code), J. Lindgren (jtlindgr{at}cs.helsinki.fi) (RBF kernel), Steven Hugg (hugg@fasterlight.com) (refactored, LRU cache), Bernhard Pfahringer (bernhard@cs.waikato.ac.nz) (full cache)

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
protected int	m_cacheHits Counts the number of kernel cache hits.
protected int	m_cacheSize The size of the cache (a prime number)
protected int	m_cacheSlots number of cache slots in an entry
protected int	m_kernelEvals Counts the number of kernel evaluations.
protected double[][]	m_kernelMatrix The kernel matrix if full cache is used (i.e.
protected long[]	m_keys
protected int	m_numInsts The number of instance in the dataset
protected double[]	m_storage Kernel cache

Fields inherited from class weka.classifiers.functions.supportVector.Kernel

m_ChecksTurnedOff, m_data, m_Debug

Constructor Summary

Constructors

Modifier	Constructor and Description
	CachedKernel() default constructor - does nothing.
protected	CachedKernel(Instances data, int cacheSize) Initializes the kernel cache.

Method Summary

Modifier and Type	Method and Description
void	buildKernel(Instances data) builds the kernel with the given data.
String	cacheSizeTipText() Returns the tip text for this property
void	clean() Frees the cache used by the kernel.
protected double	dotProd(Instance inst1, Instance inst2) Calculates a dot product between two instances
double	eval(int id1, int id2, Instance inst1) Implements the abstract function of Kernel using the cache.
protected abstract double	evaluate(int id1, int id2, Instance inst1) This method is overridden in subclasses to implement specific kernels.
int	getCacheSize() Gets the size of the cache
String[]	getOptions() Gets the current settings of the Kernel.
protected void	initVars(Instances data) initializes variables etc.
Enumeration	listOptions() Returns an enumeration describing the available options.
int	numCacheHits() Returns the number of cache hits on dot products.
int	numEvals() Returns the number of time Eval has been called.
void	setCacheSize(int value) Sets the size of the cache to use (a prime number)
void	setOptions(String[] options) Parses a given list of options.

Methods inherited from class weka.classifiers.functions.supportVector.Kernel

checksTurnedOffTipText, debugTipText, forName, getCapabilities, getChecksTurnedOff, getDebug, getRevision, globalInfo, makeCopies, makeCopy, setChecksTurnedOff, setDebug

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

m_kernelEvals

protected int m_kernelEvals

Counts the number of kernel evaluations.

m_cacheHits

protected int m_cacheHits

Counts the number of kernel cache hits.

m_cacheSize

protected int m_cacheSize

The size of the cache (a prime number)

m_storage

protected double[] m_storage

Kernel cache

m_keys

protected long[] m_keys

m_kernelMatrix

protected double[][] m_kernelMatrix

The kernel matrix if full cache is used (i.e. size is set to 0)

m_numInsts

protected int m_numInsts

The number of instance in the dataset

m_cacheSlots

protected int m_cacheSlots

number of cache slots in an entry

Constructor Detail

CachedKernel

public CachedKernel()

default constructor - does nothing.

CachedKernel

protected CachedKernel(Instances data,
                       int cacheSize)
                throws Exception

Initializes the kernel cache. The actual size of the cache in bytes is (64 * cacheSize).

Parameters:

data - the data to use

cacheSize - the cache size

Throws:

Exception - if something goes wrong

Method Detail

listOptions

public Enumeration listOptions()

Returns an enumeration describing the available options.

Specified by:

listOptions in interface OptionHandler

Overrides:

listOptions in class Kernel

Returns:

an enumeration of all the available options.

setOptions

public void setOptions(String[] options)
                throws Exception

Parses a given list of options.

Specified by:

setOptions in interface OptionHandler

Overrides:

setOptions in class Kernel

Parameters:

options - the list of options as an array of strings

Throws:

Exception - if an option is not supported

getOptions

public String[] getOptions()

Gets the current settings of the Kernel.

Specified by:

getOptions in interface OptionHandler

Overrides:

getOptions in class Kernel

Returns:

an array of strings suitable for passing to setOptions

evaluate

protected abstract double evaluate(int id1,
                                   int id2,
                                   Instance inst1)
                            throws Exception

This method is overridden in subclasses to implement specific kernels.

Parameters:

id1 - the index of instance 1

id2 - the index of instance 2

inst1 - the instance 1 object

Returns:

the dot product

Throws:

Exception - if something goes wrong

eval

public double eval(int id1,
                   int id2,
                   Instance inst1)
            throws Exception

Implements the abstract function of Kernel using the cache. This method uses the evaluate() method to do the actual dot product.

Specified by:

eval in class Kernel

Parameters:

id1 - the index of the first instance in the dataset

id2 - the index of the second instance in the dataset

inst1 - the instance corresponding to id1 (used if id1 == -1)

Returns:

the result of the kernel function

Throws:

Exception - if something goes wrong

numEvals

public int numEvals()

Returns the number of time Eval has been called.

Specified by:

numEvals in class Kernel

Returns:

the number of kernel evaluation.

numCacheHits

public int numCacheHits()

Returns the number of cache hits on dot products.

Specified by:

numCacheHits in class Kernel

Returns:

the number of cache hits.

clean

public void clean()

Frees the cache used by the kernel.

Specified by:

clean in class Kernel

dotProd

protected final double dotProd(Instance inst1,
                               Instance inst2)
                        throws Exception

Calculates a dot product between two instances

Parameters:

inst1 - the first instance

inst2 - the second instance

Returns:

the dot product of the two instances.

Throws:

Exception - if an error occurs

setCacheSize

public void setCacheSize(int value)

Sets the size of the cache to use (a prime number)

Parameters:

value - the size of the cache

getCacheSize

public int getCacheSize()

Gets the size of the cache

Returns:

the cache size

cacheSizeTipText

public String cacheSizeTipText()

Returns the tip text for this property

Returns:

tip text for this property suitable for displaying in the explorer/experimenter gui

initVars

protected void initVars(Instances data)

initializes variables etc.

Overrides:

initVars in class Kernel

Parameters:

data - the data to use

buildKernel

public void buildKernel(Instances data)
                 throws Exception

builds the kernel with the given data. Initializes the kernel cache. The actual size of the cache in bytes is (64 * cacheSize).

Overrides:

buildKernel in class Kernel

Parameters:

data - the data to base the kernel on

Throws:

Exception - if something goes wrong