it.uniroma2.sag.kelp.kernel.cache

Class StripeKernelCache

java.lang.Object

it.uniroma2.sag.kelp.kernel.cache.KernelCache

it.uniroma2.sag.kelp.kernel.cache.StripeKernelCache

All Implemented Interfaces:

Serializable

public class StripeKernelCache
extends KernelCache
implements Serializable

Given a dataset, this cache stores kernel computations in "Stripes", i.e. whole rows of the complete Gram Matrix. In other words given a subset S of the examples in the dataset D, the cache is able to store all the kernel computations between any example in S and any example in D. To reduce the requirement of memory space, the cache stores only n stripes. When the number of stripes is exceeded, they are removed according to a FIFO policy.

Author:

Danilo Croce

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
StripeKernelCache()
StripeKernelCache(Dataset dataset)
StripeKernelCache(Dataset dataset, int maxNumberOfRows)
StripeKernelCache(int maxNumberOfRows, int numberOfColumns)

Method Summary

Methods

Modifier and Type	Method and Description
void	flushCache() Empties the cache
int	getMaxNumberOfRows()
int	getNumberOfColumns()
protected Float	getStoredKernelValue(Example exA, Example exB) Retrieves in the cache the kernel operation between two examples
void	setKernelValue(Example exA, Example exB, float value) Stores a kernel computation in cache
void	setMaxNumberOfRows(int maxNumberOfRows)
void	setNumberOfColumns(int numberOfColumns)

Methods inherited from class it.uniroma2.sag.kelp.kernel.cache.KernelCache

getCacheHits, getCacheMisses, getKernelValue, resetCacheStats

Methods inherited from class java.lang.Object

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

Constructor Detail

StripeKernelCache

public StripeKernelCache(Dataset dataset)

Parameters:

dataset - The cache is initialized with a number of rows (and columns) that is equal to the size of the dataset

StripeKernelCache

public StripeKernelCache(Dataset dataset,
                 int maxNumberOfRows)

Parameters:

dataset - The cache is initialized with a number of columns that is equal to the size of the dataset

maxNumberOfItems - The maximum number of stripes in the dataset

StripeKernelCache

public StripeKernelCache(int maxNumberOfRows,
                 int numberOfColumns)

Parameters:

maxNumberOfRows - The maximum number of stripes in the matrix

numberOfColumns - The maximum number of columns in the matrix

StripeKernelCache

public StripeKernelCache()

Method Detail

getMaxNumberOfRows

public int getMaxNumberOfRows()

Returns:

the maxNumberOfRows

setMaxNumberOfRows

public void setMaxNumberOfRows(int maxNumberOfRows)

Parameters:

maxNumberOfRows - the maxNumberOfRows to set

getNumberOfColumns

public int getNumberOfColumns()

Returns:

the numberOfColumns

setNumberOfColumns

public void setNumberOfColumns(int numberOfColumns)

Parameters:

numberOfColumns - the numberOfColumns to set

setKernelValue

public void setKernelValue(Example exA,
                  Example exB,
                  float value)

Description copied from class: KernelCache

Stores a kernel computation in cache

Specified by:

setKernelValue in class KernelCache

Parameters:

exA - the first example

exB - the second example

value - the kernel value to be stored in cache

flushCache

public void flushCache()

Description copied from class: KernelCache

Empties the cache

Specified by:

flushCache in class KernelCache

getStoredKernelValue

protected Float getStoredKernelValue(Example exA,
                         Example exB)

Description copied from class: KernelCache

Retrieves in the cache the kernel operation between two examples

Specified by:

getStoredKernelValue in class KernelCache

Parameters:

exA - the first example

exB - the second example

Returns:

the kernel similarity, null if a cache miss occurs