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.ml.clustering.spectral;
031
032import java.util.Iterator;
033
034import org.apache.logging.log4j.Logger;
035import org.apache.logging.log4j.LogManager;
036import org.openimaj.ml.clustering.DataClusterer;
037import org.openimaj.ml.clustering.IndexClusters;
038import org.openimaj.ml.clustering.SpatialClusterer;
039import org.openimaj.ml.clustering.SpatialClusters;
040import org.openimaj.util.pair.DoubleObjectPair;
041import org.openimaj.util.pair.IndependentPair;
042
043import ch.akuhn.matrix.Vector;
044import ch.akuhn.matrix.Vector.Entry;
045import ch.akuhn.matrix.eigenvalues.Eigenvalues;
046
047/**
048 * For a given set of {@link Eigenvalues} perform the stages of spectral
049 * clustering which involve the selection of the best eigen values and the
050 * calling of an internal clustering algorithm
051 *
052 * @author Sina Samangooei (ss@ecs.soton.ac.uk)
053 */
054public class PreparedSpectralClustering implements DataClusterer<Eigenvalues, SpectralIndexedClusters> {
055        final static Logger logger = LogManager.getLogger(PreparedSpectralClustering.class);
056        private SpectralClusteringConf<double[]> conf;
057
058        /**
059         * @param conf
060         */
061        public PreparedSpectralClustering(SpectralClusteringConf<double[]> conf) {
062                this.conf = conf;
063        }
064
065        @Override
066        public int[][] performClustering(Eigenvalues data) {
067                return cluster(data).clusters();
068        }
069
070        @Override
071        public SpectralIndexedClusters cluster(Eigenvalues eig) {
072                // Also normalise each row
073                final IndependentPair<double[], double[][]> lowestCols = bestCols(eig);
074                // Using the eigenspace, cluster
075                return eigenspaceCluster(lowestCols);
076        }
077
078        protected SpectralIndexedClusters eigenspaceCluster(IndependentPair<double[], double[][]> lowestCols) {
079                final SpatialClusterer<? extends SpatialClusters<double[]>, double[]> clusterer = conf.internal.apply(lowestCols);
080                // Cluster the rows with the internal spatial clusterer
081                final SpatialClusters<double[]> cluster = clusterer.cluster(lowestCols.getSecondObject());
082                // if the clusters contain the cluster indexes of the training examples
083                // use those
084                if (cluster instanceof IndexClusters) {
085                        final IndexClusters clusters = new IndexClusters(((IndexClusters) cluster).clusters());
086                        // logger.debug(clusters);
087                        return new SpectralIndexedClusters(clusters, lowestCols);
088                }
089                // Otherwise attempt to assign values to clusters
090                final int[] clustered = cluster.defaultHardAssigner().assign(lowestCols.getSecondObject());
091                // done!
092                return new SpectralIndexedClusters(new IndexClusters(clustered), lowestCols);
093        }
094
095        protected IndependentPair<double[], double[][]> bestCols(final Eigenvalues eig) {
096
097                int eigenVectorSelect = conf.eigenChooser.nEigenVectors(this.conf.laplacian.eigenIterator(eig), eig.getN());
098                final int eigenVectorSkip = this.conf.skipEigenVectors;
099                logger.debug("Selected dimensions: " + eigenVectorSelect);
100                logger.debug("Skipping dimesions: " + eigenVectorSkip);
101                eigenVectorSelect -= eigenVectorSkip;
102
103                final int nrows = eig.vector[0].size();
104                final double[][] ret = new double[nrows][eigenVectorSelect];
105                final double[] retSum = new double[nrows];
106                final double[] eigvals = new double[eigenVectorSelect];
107                final Iterator<DoubleObjectPair<Vector>> iterator = this.conf.laplacian.eigenIterator(eig);
108                // Skip a few at the beggining
109                for (int i = 0; i < eigenVectorSkip; i++)
110                        iterator.next();
111                int col = 0;
112                // Calculate U matrix (containing n smallests eigen valued columns)
113                for (; iterator.hasNext();) {
114                        final DoubleObjectPair<Vector> v = iterator.next();
115                        eigvals[col] = v.first;
116
117                        for (final Entry d : v.second.entries()) {
118                                double elColI = d.value;
119                                if (conf.eigenValueScale) {
120                                        elColI *= Math.sqrt(v.first);
121                                }
122                                ret[d.index][col] = elColI;
123                                retSum[d.index] += elColI * elColI;
124                        }
125                        col++;
126                        if (col == eigenVectorSelect)
127                                break;
128                }
129
130                if (!conf.eigenValueScale) {
131                        // normalise rows
132                        for (int i = 0; i < ret.length; i++) {
133                                final double[] row = ret[i];
134                                for (int j = 0; j < row.length; j++) {
135                                        row[j] /= Math.sqrt(retSum[i]);
136                                }
137                        }
138                }
139
140                return IndependentPair.pair(eigvals, ret);
141        }
142
143}