001/**
002 * Copyright (c) 2011, The University of Southampton and the individual contributors.
003 * All rights reserved.
004 *
005 * Redistribution and use in source and binary forms, with or without modification,
006 * are permitted provided that the following conditions are met:
007 *
008 *   *  Redistributions of source code must retain the above copyright notice,
009 *      this list of conditions and the following disclaimer.
010 *
011 *   *  Redistributions in binary form must reproduce the above copyright notice,
012 *      this list of conditions and the following disclaimer in the documentation
013 *      and/or other materials provided with the distribution.
014 *
015 *   *  Neither the name of the University of Southampton nor the names of its
016 *      contributors may be used to endorse or promote products derived from this
017 *      software without specific prior written permission.
018 *
019 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
020 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
021 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
022 * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
023 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
024 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
025 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
026 * ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
027 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
028 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
029 */
030package org.openimaj.math.model;
031
032import gov.sandia.cognition.learning.algorithm.bayes.VectorNaiveBayesCategorizer;
033import gov.sandia.cognition.learning.data.DefaultInputOutputPair;
034import gov.sandia.cognition.learning.data.InputOutputPair;
035import gov.sandia.cognition.math.matrix.Vector;
036import gov.sandia.cognition.math.matrix.VectorFactory;
037import gov.sandia.cognition.statistics.distribution.UnivariateGaussian.PDF;
038
039import java.util.ArrayList;
040import java.util.List;
041
042import org.openimaj.util.pair.IndependentPair;
043
044/**
045 * An implementation of a {@link EstimatableModel} that uses a
046 * {@link VectorNaiveBayesCategorizer} to associate vectors (actually double[])
047 * with a category based on the naive bayes model.
048 * 
049 * @author Jonathon Hare (jsh2@ecs.soton.ac.uk)
050 * 
051 * @param <T>
052 *            The type of class/category predicted by the model
053 */
054public class GaussianVectorNaiveBayesModel<T> implements EstimatableModel<double[], T> {
055        VectorNaiveBayesCategorizer.BatchGaussianLearner<T> learner = new VectorNaiveBayesCategorizer.BatchGaussianLearner<T>();
056        private VectorNaiveBayesCategorizer<T, PDF> model;
057
058        @Override
059        public boolean estimate(List<? extends IndependentPair<double[], T>> data) {
060                final List<InputOutputPair<Vector, T>> cfdata = new ArrayList<InputOutputPair<Vector, T>>();
061
062                for (final IndependentPair<double[], T> d : data) {
063                        final InputOutputPair<Vector, T> iop = new DefaultInputOutputPair<Vector, T>(VectorFactory.getDefault()
064                                        .copyArray(d.firstObject()), d.secondObject());
065                        cfdata.add(iop);
066                }
067
068                model = learner.learn(cfdata);
069
070                return true;
071        }
072
073        @Override
074        public T predict(double[] data) {
075                return model.evaluate(VectorFactory.getDefault().copyArray(data));
076        }
077
078        @Override
079        public int numItemsToEstimate() {
080                return 0;
081        }
082
083        @Override
084        @SuppressWarnings("unchecked")
085        public GaussianVectorNaiveBayesModel<T> clone() {
086                try {
087                        return (GaussianVectorNaiveBayesModel<T>) super.clone();
088                } catch (final CloneNotSupportedException e) {
089                        throw new RuntimeException(e);
090                }
091        }
092
093        /**
094         * Testing
095         * 
096         * @param args
097         */
098        public static void main(String[] args) {
099                final GaussianVectorNaiveBayesModel<Boolean> model = new GaussianVectorNaiveBayesModel<Boolean>();
100
101                final List<IndependentPair<double[], Boolean>> data = new ArrayList<IndependentPair<double[], Boolean>>();
102
103                data.add(IndependentPair.pair(new double[] { 0 }, true));
104                data.add(IndependentPair.pair(new double[] { 0.1 }, true));
105                data.add(IndependentPair.pair(new double[] { -0.1 }, true));
106
107                data.add(IndependentPair.pair(new double[] { 9.9 }, false));
108                data.add(IndependentPair.pair(new double[] { 10 }, false));
109                data.add(IndependentPair.pair(new double[] { 10.1 }, false));
110
111                model.estimate(data);
112
113                final double[] q = new double[] { 5.0 };
114
115                System.out.println(model.predict(q));
116
117                System.out.println(" logP(true): "
118                                + model.model.computeLogPosterior(VectorFactory.getDefault().copyArray(q), true));
119                System.out.println("logP(false): "
120                                + model.model.computeLogPosterior(VectorFactory.getDefault().copyArray(q), false));
121
122                System.out.println("    P(true): " + model.model.computePosterior(VectorFactory.getDefault().copyArray(q), true));
123                System.out
124                                .println("   P(false): " + model.model.computePosterior(VectorFactory.getDefault().copyArray(q), false));
125
126                System.out.println(model.model.getPriors());
127        }
128}