bncinference.java

bayes network classifier toolbox 贝叶斯网络分类工具箱

/**
 *  JBNC - Bayesian Network Classifiers Toolbox <p>
 *
 *  Latest release available at http://sourceforge.net/projects/jbnc/ <p>
 *
 *  Copyright (C) 1999-2003 Jarek Sacha <p>
 *
 *  This program 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 2 of the License, or (at your option)
 *  any later version. <p>
 *
 *  This program 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. <p>
 *
 *  You should have received a copy of the GNU General Public License along with
 *  this program; if not, write to the Free Software Foundation, Inc., 59 Temple
 *  Place - Suite 330, Boston, MA 02111-1307, USA. <br>
 *  http://www.fsf.org/licenses/gpl.txt
 */
package jbnc.graphs;

import BayesianNetworks.BayesNet;
import BayesianNetworks.DiscreteVariable;
import BayesianNetworks.ProbabilityFunction;
import BayesianNetworks.ProbabilityVariable;

import java.io.Serializable;

/**
 * Efficient calculation of the joint probability distribution and and conditional probability of the class variable
 * given values of all attributes. Assumes that all nodes are instantiated.
 *
 * @author Jarek Sacha
 * @since June 1, 1999
 */
public class BNCInference implements Serializable {

    /** */
    protected int nbVars = 0;

    /** */
    protected int nbClasses = 0;

    /**
     * Probability function values for each of the network nodes.
     */
    protected double[][] probFuncs = null;

    /**
     * Variables for each of the probability function.
     */
    protected int[][] variables;

    /**
     * Used to calculate index within the probFunction.
     */
    protected int[][] radix;

    /** */
    protected double[] condClassProb = null;


    /**
     * @param net Description of Parameter
     * @throws Exception Description of Exception
     */
    public BNCInference(BayesNet net) throws Exception {
        ProbabilityFunction[] pFuncs = net.get_probability_functions();
        ProbabilityVariable[] pVars = net.get_probability_variables();

        if (pFuncs.length != pVars.length) {
            throw new Exception("Number of functions does not " +
                    "match number of variables in the network.");
        }

        nbVars = pVars.length;

        int[] varSize = new int[nbVars];
        for (int i = 0; i < nbVars; ++i) {
            varSize[i] = pVars[i].number_values();
        }

        nbClasses = varSize[nbVars - 1];
        condClassProb = new double[nbClasses];

        probFuncs = new double[nbVars][];
        variables = new int[nbVars][];
        radix = new int[nbVars][];
        for (int i = 0; i < nbVars; ++i) {
            DiscreteVariable[] vars = pFuncs[i].get_variables();
            int thisIndex = vars[0].get_index();

            probFuncs[thisIndex] = pFuncs[i].get_values();
            variables[thisIndex] = pFuncs[i].get_indexes();

            int[] varCycle = new int[variables[thisIndex].length];
            varCycle[varCycle.length - 1] = 1;
            for (int j = varCycle.length - 2; j >= 0; --j) {
                varCycle[j] = varCycle[j + 1] * varSize[variables[thisIndex][j + 1]];
            }

            radix[thisIndex] = varCycle;
        }

    }


    /**
     * Return the value of the joint probability for the instantiation of the variables given by parameter 'vars'.
     *
     * @param aCase Values for each of the varaibles in the network. The last variable is class variable.
     * @return The JointProb value
     * @throws Exception when size of array 'aCase' is different then expected.
     */
    public double getJointProb(int[] aCase) throws Exception {
        if (nbVars != aCase.length) {
            throw new Exception("Incorrect number of variables in the case. "
                    + "Expecting " + nbVars + ", got " + aCase.length);
        }

        double prob = 1;

        for (int i = 0; i < nbVars; ++i) {
            // Claculate index of the probability function.
            int probIndex = 0;
            for (int j = 0; j < variables[i].length; ++j) {
                int r = radix[i][j];
                int v = variables[i][j];
                int c = aCase[v];
                probIndex += r * c;
//        probIndex = radix[i][j]*aCase[variables[i][j]];
            }

            prob *= probFuncs[i][probIndex];
        }

        return prob;
    }


    /**
     * Return values of the class node conditional probability given instantiation of attribute variables.
     *
     * @param aCase Values for each of the varaibles in the network. The last variable is class variable. Its value is
     *              ignored.
     * @return The CondClassProb value
     * @throws Exception when size of array 'aCase' is different then expected.
     */
    public double[] getCondClassProb(int[] aCase) throws Exception {
        if (nbVars != aCase.length) {
            throw new Exception("Incorrect number of variables in the case. "
                    + "Expecting " + nbVars + ", got " + aCase.length);
        }

        int origClassIndex = aCase[nbVars - 1];

        double pA = 0;
        // Mariginal prtobability of the attribute variables
        for (int c = 0; c < nbClasses; ++c) {
            aCase[nbVars - 1] = c;
            condClassProb[c] = getJointProb(aCase);
            pA += condClassProb[c];
        }

        // Normalize
        for (int c = 0; c < nbClasses; ++c) {
            condClassProb[c] /= pA;
        }

        // Restore the modified input parameter
        aCase[nbVars - 1] = origClassIndex;

        return condClassProb;
    }

}