package jsat.distributions.kernels;
import java.util.*;
import jsat.DataSet;
import jsat.distributions.Distribution;
import jsat.distributions.LogUniform;
import jsat.distributions.discrete.UniformDiscrete;
import jsat.linear.Vec;
import jsat.parameters.DoubleParameter;
import jsat.parameters.Parameter;
/**
* Provides a Polynomial Kernel of the form
* k(x,y) = (alpha * x.y + c)^d
* @author Edward Raff
*/
public class PolynomialKernel extends BaseKernelTrick
{
private static final long serialVersionUID = 9123109691782934745L;
private double degree;
private double alpha;
private double c;
/**
* Creates a new polynomial kernel
* @param degree the degree of the polynomial
* @param alpha the term to scale the dot product by
* @param c the additive term
*/
public PolynomialKernel(double degree, double alpha, double c)
{
this.degree = degree;
this.alpha = alpha;
this.c = c;
}
/**
* Defaults alpha = 1 and c = 1
* @param degree the degree of the polynomial
*/
public PolynomialKernel(double degree)
{
this(degree, 1, 1);
}
/**
* Sets the scaling factor for the dot product, this is equivalent to
* multiplying each value in the data set by a constant factor
* @param alpha the scaling factor
*/
public void setAlpha(double alpha)
{
if(Double.isInfinite(alpha) || Double.isNaN(alpha) || alpha == 0)
throw new IllegalArgumentException("alpha must be a real non zero value, not " + alpha);
this.alpha = alpha;
}
/**
* 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
* {@link #setAlpha(double) alpha}.
* @param c the non negative additive term
*/
public void setC(double c)
{
if(c < 0 || Double.isNaN(c) || Double.isInfinite(c))
throw new IllegalArgumentException("C must be non negative, not " + c);
this.c = c;
}
/**
* Sets the degree of the polynomial
* @param d the degree of the polynomial
*/
public void setDegree(double d)
{
this.degree = d;
}
/**
* Returns the scaling parameter
* @return the scaling parameter
*/
public double getAlpha()
{
return alpha;
}
/**
* Returns the additive constant
* @return the additive constant
*/
public double getC()
{
return c;
}
/**
* Returns the degree of the polynomial
* @return the degree of the polynomial
*/
public double getDegree()
{
return degree;
}
@Override
public double eval(Vec a, Vec b)
{
return Math.pow(c+a.dot(b)*alpha, degree);
}
@Override
public String toString()
{
return "Polynomial Kernel ( degree="+degree + ", c=" + c + ", alpha=" + alpha + ")";
}
/**
* Guesses the distribution to use for the degree parameter
*
* @param d the dataset to get the guess for
* @return the guess for the degree parameter
* @see #setDegree(double)
*/
public static Distribution guessDegree(DataSet d)
{
return new UniformDiscrete(2, 9);
}
@Override
public PolynomialKernel clone()
{
return new PolynomialKernel(degree, alpha, c);
}
}
Ads help maintain this website.