jsat.distributions.kernels

Class SigmoidKernel

java.lang.Object

jsat.distributions.kernels.BaseKernelTrick

jsat.distributions.kernels.SigmoidKernel

All Implemented Interfaces:

java.io.Serializable, java.lang.Cloneable, KernelTrick, Parameterized

public class SigmoidKernel
extends BaseKernelTrick

Provides an implementation of the Sigmoid (Hyperbolic Tangent) Kernel, which is of the form:
k(x, y) = tanh(alpha * < x, y > +c)
Technically, this kernel is not positive definite.

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
SigmoidKernel(double alpha) Creates a new Sigmoid Kernel with a bias term of 1
SigmoidKernel(double alpha, double C) Creates a new Sigmoid Kernel

Method Summary

Modifier and Type	Method and Description
SigmoidKernel	clone()
double	eval(Vec a, Vec b) Evaluate this kernel function for the two given vectors.
double	getAlpha() Returns the scaling parameter
double	getC() Returns the additive constant
static Distribution	guessAlpha(DataSet d) Guesses a distribution for the α parameter
static Distribution	guessC(DataSet d) Guesses a distribution for the α parameter
void	setAlpha(double alpha) Sets the scaling factor for the dot product, this is equivalent to multiplying each value in the data set by a constant factor
void	setC(double c) Sets the additive term, when set to one this is equivalent to adding a bias term of 1 to each vector.

Methods inherited from class jsat.distributions.kernels.BaseKernelTrick

addToCache, eval, eval, evalSum, evalSum, getAccelerationCache, getQueryInfo, normalized, supportsAcceleration

Methods inherited from class java.lang.Object

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

Methods inherited from interface jsat.distributions.kernels.KernelTrick

toString

Methods inherited from interface jsat.parameters.Parameterized

getParameter, getParameters

Constructor Detail

SigmoidKernel

public SigmoidKernel(double alpha,
                     double C)

Creates a new Sigmoid Kernel

Parameters:

alpha - the scaling factor for the dot product

C - the additive constant

SigmoidKernel

public SigmoidKernel(double alpha)

Creates a new Sigmoid Kernel with a bias term of 1

Parameters:

alpha - the scaling factor for the dot product

Method Detail

setAlpha

public void setAlpha(double alpha)

Sets the scaling factor for the dot product, this is equivalent to multiplying each value in the data set by a constant factor

Parameters:

alpha - the scaling factor

getAlpha

public double getAlpha()

Returns the scaling parameter

Returns:

the scaling parameter

setC

public void setC(double c)

Sets the additive term, when set to one this is equivalent to adding a bias term of 1 to each vector. This is done after the scaling by alpha.

Parameters:

c - the non negative additive term

getC

public double getC()

Returns the additive constant

Returns:

the additive constant

eval

public double eval(Vec a,
                   Vec b)

Description copied from interface: KernelTrick

Evaluate this kernel function for the two given vectors.

Parameters:

a - the first vector

b - the first vector

Returns:

the evaluation

guessAlpha

public static Distribution guessAlpha(DataSet d)

Guesses a distribution for the α parameter

Parameters:

d - the data to get the guess for

Returns:

a distribution for the α parameter

guessC

public static Distribution guessC(DataSet d)

Guesses a distribution for the α parameter

Parameters:

d - the data to get the guess for

Returns:

a distribution for the α parameter

clone

public SigmoidKernel clone()

Specified by:

clone in interface KernelTrick

Specified by:

clone in class BaseKernelTrick