fr.lip6.kernel.adaptative
Class WeightedSumKernel<T>

java.lang.Object
  fr.lip6.kernel.Kernel<T>
      fr.lip6.kernel.adaptative.WeightedSumKernel<T>

Type Parameters:

T -

All Implemented Interfaces:

java.io.Serializable

public class WeightedSumKernel<T>
extends Kernel<T>

Major kernel computed as a weighted sum of minor kernels : K = w_i * k_i

Author:

dpicard

See Also:

Serialized Form

Field Summary

Fields inherited from class fr.lip6.kernel.Kernel

name

Constructor Summary

WeightedSumKernel()

WeightedSumKernel(java.util.Hashtable<Kernel<T>,java.lang.Double> h)
Sets the weights to h.

Method Summary

void	addKernel(Kernel<T> k) adds a kernel to the sum with weight 1.0
void	addKernel(Kernel<T> k, double d) adds a kernel to the sum with weight d
double[][]	getKernelMatrix(java.util.ArrayList<TrainingSample<T>> e) return the Gram Matrix of this kernel computed on given samples
double	getWeight(Kernel<T> k) gets the weights of kernel k
java.util.Hashtable<Kernel<T>,java.lang.Double>	getWeights() get the list of kernels and associated weights.
void	removeKernel(Kernel<T> k) removes kernel k from the sum
void	setWeight(Kernel<T> k, java.lang.Double d) Sets the weight of kernel k
double	valueOf(T t1) kernel similarity to zero
double	valueOf(T t1, T t2) compute the kernel similarity between two element of input space

Methods inherited from class fr.lip6.kernel.Kernel

getNormalizedKernelMatrix, normalizedValueOf, setName, toString

Methods inherited from class java.lang.Object

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

Constructor Detail

WeightedSumKernel

public WeightedSumKernel()

WeightedSumKernel

public WeightedSumKernel(java.util.Hashtable<Kernel<T>,java.lang.Double> h)

Sets the weights to h. Beware! It does not make a copy of h!

Parameters:

h -

Method Detail

addKernel

public void addKernel(Kernel<T> k)

adds a kernel to the sum with weight 1.0

Parameters:

k -

addKernel

public void addKernel(Kernel<T> k,
                      double d)

adds a kernel to the sum with weight d

Parameters:

k -

d -

removeKernel

public void removeKernel(Kernel<T> k)

removes kernel k from the sum

Parameters:

k -

getWeight

public double getWeight(Kernel<T> k)

gets the weights of kernel k

Parameters:

k -

Returns:

the weight associated with k

setWeight

public void setWeight(Kernel<T> k,
                      java.lang.Double d)

Sets the weight of kernel k

Parameters:

k -

d -

valueOf

public double valueOf(T t1,
                      T t2)

Description copied from class: Kernel

compute the kernel similarity between two element of input space

Specified by:

valueOf in class Kernel<T>

Parameters:

t1 - first element

t2 - second element

Returns:

the kernel value

valueOf

public double valueOf(T t1)

Description copied from class: Kernel

kernel similarity to zero

Specified by:

valueOf in class Kernel<T>

Parameters:

t1 - the element to compute the similarity to zero

getWeights

public java.util.Hashtable<Kernel<T>,java.lang.Double> getWeights()

get the list of kernels and associated weights.

Returns:

hashtable containing kernels as keys and weights as values.

getKernelMatrix

public double[][] getKernelMatrix(java.util.ArrayList<TrainingSample<T>> e)

Description copied from class: Kernel

return the Gram Matrix of this kernel computed on given samples

Overrides:

getKernelMatrix in class Kernel<T>

Returns:

double[][] containing similarities in the order of the e.