bham.leakiest.infotheory

Class KernelFunction

java.lang.Object

bham.leakiest.infotheory.KernelFunction

public class KernelFunction
extends java.lang.Object

This file contains the method for estimating a continuous probability distribution from sampled data using a kernel function.

Version:

1.4.6

Author:

Tom Chothia, Yusuke Kawamoto

Field Summary

Fields

Modifier and Type	Field and Description
static int	resolutionForMI Parameter to approximate the mutual information for the continuous data

Constructor Summary

Constructors

Constructor and Description
KernelFunction(java.util.ArrayList<double[]> observedDataList) Initialises a kernel function with a list of observation data.

Method Summary

Modifier and Type	Method and Description
double	calcContinuousApproxMI(double[] distInput, java.util.ArrayList<double[]> Datalist) Calculates the estimated mutual information.
double[]	probInputDist(java.util.ArrayList<double[]> DataList) Calculates estimated input distribution.

Methods inherited from class java.lang.Object

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

Field Detail

resolutionForMI

public static int resolutionForMI

Parameter to approximate the mutual information for the continuous data

Constructor Detail

KernelFunction

public KernelFunction(java.util.ArrayList<double[]> observedDataList)

Initialises a kernel function with a list of observation data.

Parameters:

observedDataList - list of observation data

Method Detail

calcContinuousApproxMI

public double calcContinuousApproxMI(double[] distInput,
                                     java.util.ArrayList<double[]> Datalist)

Calculates the estimated mutual information. This method returns:
I(Q,W) = I (X;Y) = Σ_x ∫_y Q(x).W(y|x)log(W(y|x)/Y(y)) dy = Q(input_0) ∫_y W(y|input_0)log(W(y|input_0)/Y(y)) dy + ... + Q(input_k) ∫_y W(y|input_k)log(W(y|input_k)/Y(y)) dy
Y(y) = Q(input_0) W(y|input_0) + ... + Q(input_k) W(y|input_k)
∫_y W(y|input_x)log(W(y|input_x)/Y(y)) dy = ∫_n^{n_1} W(y|input_x)log(W(y|input_x)/Y(y)) dy + ∫_{n_1}^{n_2} W(y|input_x)log(W(y|input_x)/Y(y)) dy + ...
where ∫_{n_1}^{n_2} W(y|input_x)log(W(y|input_x)/Y(y)) dy ~ W(n_1|input_x)log(W(n_1|input_x)/Y(n_1))*(n_2-n_1)

Parameters:

distInput - input probability distribution

Datalist - list of observed data

Returns:

estimated mutual information

probInputDist

public double[] probInputDist(java.util.ArrayList<double[]> DataList)

Calculates estimated input distribution.

Parameters:

DataList - list of observation data

Returns:

estimated input distribution