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.pgm.vb.lda.mle;
031
032import org.openimaj.pgm.util.Corpus;
033
034import cern.jet.random.engine.MersenneTwister;
035
036/**
037 * Initialisation strategies for the beta matrix in the maximum liklihood LDA.
038 * Specifically implementors are expected to initialise the sufficient statistics
039 * of beta (i.e. topicWord and topicTotal s.t. Beta_ij = topicWord_ij / topicTotal_i
040 * @author Sina Samangooei (ss@ecs.soton.ac.uk)
041 *
042 */
043public interface LDABetaInitStrategy{
044        /**
045         * Given a model and the corpus initialise the model's sufficient statistics
046         * @param model
047         * @param corpus
048         */
049        public void initModel(LDAModel model, Corpus corpus);
050        
051        /**
052         * initialises beta randomly s.t. each each topicWord >= 1 and < 2
053         * @author Sina Samangooei (ss@ecs.soton.ac.uk)
054         *
055         */
056        public static class RandomBetaInit implements LDABetaInitStrategy{
057                private MersenneTwister random;
058                
059                /**
060                 * unseeded random
061                 */
062                public RandomBetaInit() {
063                        random = new MersenneTwister();
064                }
065                
066                /**
067                 * seeded random
068                 * @param seed
069                 */
070                public RandomBetaInit(int seed) {
071                        random = new MersenneTwister(seed);
072                }
073                @Override
074                public void initModel(LDAModel model, Corpus corpus) {
075                        for (int topicIndex = 0; topicIndex < model.ntopics; topicIndex++) {
076                                for (int wordIndex = 0; wordIndex < corpus.vocabularySize(); wordIndex++) {
077                                        double topicWord = 1 + random.nextDouble();
078                                        model.incTopicWord(topicIndex,wordIndex,topicWord);
079                                        model.incTopicTotal(topicIndex, topicWord);
080                                }
081                        }
082                }
083        }
084}