com.davidbracewell.ml.sequence.hmm.MLFirstOrderHMMLearner.java Source code

Introduction

Here is the source code for com.davidbracewell.ml.sequence.hmm.MLFirstOrderHMMLearner.java

Source

/*
 * (c) 2005 David B. Bracewell
 *
 * Licensed to the Apache Software Foundation (ASF) under one
 * or more contributor license agreements.  See the NOTICE file
 * distributed with this work for additional information
 * regarding copyright ownership.  The ASF licenses this file
 * to you under the Apache License, Version 2.0 (the
 * "License"); you may not use this file except in compliance
 * with the License.  You may obtain a copy of the License at
 *
 *   http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing,
 * software distributed under the License is distributed on an
 * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
 * KIND, either express or implied.  See the License for the
 * specific language governing permissions and limitations
 * under the License.
 */

package com.davidbracewell.ml.sequence.hmm;

import com.davidbracewell.collection.CollectionUtils;
import com.davidbracewell.logging.Logger;
import com.davidbracewell.math.DoubleEntry;
import com.davidbracewell.math.distribution.DiscreteEstimator;
import com.davidbracewell.math.distribution.conditional.DiscreteConditionalEstimator;
import com.davidbracewell.ml.Instance;
import com.davidbracewell.ml.sequence.SequenceLearner;
import com.davidbracewell.ml.sequence.extractor.ContextExtractorSet;
import com.davidbracewell.ml.sequence.indexers.DataIndexer;
import com.google.common.base.Stopwatch;
import com.google.common.collect.Maps;
import org.apache.mahout.math.map.OpenIntObjectHashMap;

import java.util.Iterator;
import java.util.List;
import java.util.Map;

/**
 * The type ML first order hMM learner.
 *
 * @author David B. Bracewell
 */
public class MLFirstOrderHMMLearner<V> extends SequenceLearner<V, FirstOrderHMM<V>> {
    private static final long serialVersionUID = 1753326911357058904L;
    private static final Logger log = Logger.getLogger(MLFirstOrderHMMLearner.class);
    DiscreteEstimator pi;
    DiscreteConditionalEstimator alpha;
    Map<Integer, DiscreteConditionalEstimator> beta;

    /**
     * Instantiates a new maximum likelihood  first order hidden markov model learner.
     *
     * @param featureExtractors the feature extractors
     * @param dataIndexer       The data indexer
     */
    public MLFirstOrderHMMLearner(ContextExtractorSet<V> featureExtractors, DataIndexer<V> dataIndexer) {
        super(featureExtractors, dataIndexer);
        this.model.setContextExtractors(featureExtractors);
    }

    private void incBeta(FirstOrderHMM<V> model, Instance instance,
            Map<Integer, DiscreteConditionalEstimator> beta) {
        for (DoubleEntry entry : CollectionUtils.asIterable(instance.nonZeroIterator())) {
            if (!beta.containsKey(entry.index)) {
                beta.put(entry.index, new DiscreteConditionalEstimator(DiscreteEstimator.mleSupplier()));
            }
            beta.get(entry.index).increment(instance.getTargetValue(), entry.value);
        }
    }

    @Override
    protected void trainAllInstances(List<List<Instance>> instances) {
        log.info("Beginning training [{0} examples] ...", instances.size());
        Stopwatch sw = Stopwatch.createStarted();
        for (Iterator<List<Instance>> iterator = instances.iterator(); iterator.hasNext();) {
            trainOneSequence(iterator.next());
            iterator.remove();
        }
        sw.stop();
        log.info("Finished training in {0}", sw);
    }

    @Override
    protected void trainOneSequence(List<Instance> sequence) {
        pi.increment(sequence.get(0).getTargetValue());
        incBeta(model, sequence.get(0), beta);
        double last = sequence.get(0).getTargetValue();
        for (int i = 1; i < sequence.size(); i++) {
            double cv = sequence.get(i).getTargetValue();
            alpha.increment(last, cv);
            last = cv;
            incBeta(model, sequence.get(i), beta);
        }
    }

    @Override
    public FirstOrderHMM<V> buildModel() {
        int NF = model.getFeatures().size();
        model.pi = pi.estimate();
        model.alpha = alpha.estimate();
        model.beta = new OpenIntObjectHashMap<>();
        for (int i = 0; i < NF; i++) {
            if (beta.containsKey(i)) {
                model.beta.put(i, beta.get(i).estimate());
            }
        }
        return model;
    }

    @Override
    public void clear() {
        pi = DiscreteEstimator.laplaceEstimator();
        alpha = new DiscreteConditionalEstimator(DiscreteEstimator.mleSupplier());
        beta = Maps.newHashMap();
        model = new FirstOrderHMM<>();
    }

}//END OF MLFirstOrderHMMLearner