Package fr.lip6.jkernelmachines.kernel.adaptative

Source Code of fr.lip6.jkernelmachines.kernel.adaptative.ThreadedProductKernel

/**
    This file is part of JkernelMachines.

    JkernelMachines is free software: you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
    the Free Software Foundation, either version 3 of the License, or
    (at your option) any later version.

    JkernelMachines is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License
    along with JkernelMachines.  If not, see <http://www.gnu.org/licenses/>.

    Copyright David Picard - 2010

*/
package fr.lip6.jkernelmachines.kernel.adaptative;

import java.util.Arrays;
import java.util.HashMap;
import java.util.List;
import java.util.Map;

import fr.lip6.jkernelmachines.kernel.Kernel;
import fr.lip6.jkernelmachines.threading.ThreadedMatrixOperator;
import fr.lip6.jkernelmachines.type.TrainingSample;

/**
* Major kernel computed as a weighted product of minor kernels :
* K = k_i^{w_i}<br />
* Computation of the kernel matrix is done by running a thread on sub matrices.
* The number of threads is chosen as function of the number of available cpus.
* @author dpicard
*
* @param <T>
*/
public class ThreadedProductKernel<T> extends Kernel<T> {

  /**
   *
   */
  private static final long serialVersionUID = 7780445301175174296L;
 
 
  private Map<Kernel<T>, Double> kernels;
  protected int numThread = 0;
 
  public ThreadedProductKernel()
  {
    kernels = new HashMap<Kernel<T>, Double>();
  }

  /**
   * Sets the weights to h.
   * @param h
   */
  public ThreadedProductKernel(Map<Kernel<T>, Double> h)
  {
    kernels = new HashMap<Kernel<T>, Double>();
    kernels.putAll(h);
  }
 
  /**
   * adds a kernel to the sum with weight 1.0
   * @param k
   */
  public void addKernel(Kernel<T> k)
  {
    synchronized(kernels)
    {
      kernels.put(k, 1.0);
    }
  }
 
  /**
   * adds a kernel to the sum with weight d
   * @param k
   * @param d
   */
  public void addKernel(Kernel<T> k , double d)
  {
    synchronized(kernels)
    {
      kernels.put(k, d);
    }
  }
 
  /**
   * removes kernel k from the sum
   * @param k
   */
  public void removeKernel(Kernel<T> k)
  {
    synchronized(kernels)
    {
      kernels.remove(k);
    }
  }
 
  /**
   * gets the weights of kernel k
   * @param k
   * @return the weight associated with k
   */
  public double getWeight(Kernel<T> k)
  {
    synchronized(kernels)
    {
      Double d = kernels.get(k);
      if(d == null)
        return 0.;
      return d.doubleValue();
    }
  }
 
  /**
   * Sets the weight of kernel k
   * @param k
   * @param d
   */
  public void setWeight(Kernel<T> k, Double d)
  {
    synchronized(kernels)
    {
      kernels.put(k, d);
    }
  }
 
  @Override
  public double valueOf(T t1, T t2) {
    double sum = 1.;
    for(Kernel<T> k : kernels.keySet())
    {
      double w = kernels.get(k);
      if(w != 0)
        sum *= Math.pow(k.valueOf(t1, t2), kernels.get(k));
    }
   
    return sum;
  }

  @Override
  public double valueOf(T t1) {
    return valueOf(t1, t1);
  }
 
  /**
   * get the list of kernels and associated weights.
   * @return hashtable containing kernels as keys and weights as values.
   */
  public Map<Kernel<T>, Double> getWeights()
  {
    return kernels;
  }
 
  @Override
  public double[][] getKernelMatrix(List<TrainingSample<T>> list)
  {
    final List<TrainingSample<T>> l = list;
    //init matrix with ones
    double matrix[][] = new double[l.size()][l.size()];
    for(double[] lines : matrix)
      Arrays.fill(lines, 1.);
   

    for(final Kernel<T> k : kernels.keySet())
    {
      final double w = kernels.get(k);
     
      //check w
      if(w == 0)
        continue;
     

      final double[][] m = k.getKernelMatrix(l);
      // specific factory
      ThreadedMatrixOperator tmo = new ThreadedMatrixOperator(){
       
        @Override
        public void doLines(double[][] matrix , int from , int to) {
         
          for(int index = from ; index < to ; index++)
          {
            for(int i = 0 ; i < m[index].length ; i++)
            {
              matrix[index][i] *= Math.pow(m[index][i], w);
            }
          }
        };
       
      };
     
      matrix = tmo.getMatrix(matrix);
    }
   
    return matrix;
  }
}
TOP

Related Classes of fr.lip6.jkernelmachines.kernel.adaptative.ThreadedProductKernel

TOP
Copyright © 2018 www.massapi.com. All rights reserved.
All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.