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.dbscan;
031
032import java.util.ArrayList;
033import java.util.Iterator;
034import java.util.List;
035
036import org.openimaj.data.DataSource;
037import org.openimaj.knn.DoubleNearestNeighbours;
038import org.openimaj.knn.DoubleNearestNeighboursExact;
039import org.openimaj.knn.NearestNeighboursFactory;
040import org.openimaj.ml.clustering.DataClusterer;
041import org.openimaj.ml.clustering.SpatialClusterer;
042import org.openimaj.ml.clustering.dbscan.neighbourhood.RegionMode;
043import org.openimaj.util.pair.IntDoublePair;
044
045/**
046 * Implementation of DBSCAN (http://en.wikipedia.org/wiki/DBSCAN) using
047 * a
048 * @author Sina Samangooei (ss@ecs.soton.ac.uk)
049 *
050 */
051public class DoubleNNDBSCAN extends DBSCAN implements SpatialClusterer<DoubleDBSCANClusters, double[]>, DataClusterer<double[][], DoubleDBSCANClusters>{
052
053        private NearestNeighboursFactory<? extends DoubleNearestNeighbours, double[]> nnf;
054        private double eps;
055        private int minPts;
056        
057
058        /**
059         * Perform a DBScane with this configuration
060         * @param eps 
061         * @param minPts 
062         * @param nnf 
063         */
064        public DoubleNNDBSCAN(double eps, int minPts, NearestNeighboursFactory<? extends DoubleNearestNeighbours, double[]> nnf) {
065                this.eps = eps;
066                this.nnf = nnf;
067                this.minPts = minPts;
068        }
069        /**
070         * @param eps
071         * @param minPts
072         */
073        public DoubleNNDBSCAN(double eps, int minPts) {
074                this(eps,minPts,new DoubleNearestNeighboursExact.Factory());
075        }
076        class NNRegionMode implements RegionMode<IntDoublePair>{
077                double[][] data;
078                DoubleNearestNeighbours nn;
079                public NNRegionMode(double[][] data) {
080                        this.data = data;
081                        this.nn = nnf.create(data);
082                }
083                @Override
084                public List<IntDoublePair> regionQuery(int index) {
085                        List<IntDoublePair> res = nn.searchKNN(data[index], data.length);
086                        List<IntDoublePair> ret = new ArrayList<IntDoublePair>();
087                        for (IntDoublePair intFloatPair : res) {
088                                if(intFloatPair.second<eps)ret.add(intFloatPair);
089                                else break;
090                        }
091                        return ret;
092                }
093                
094                @Override
095                public boolean validRegion(List<IntDoublePair> region) {
096                        return region.size() >= minPts;
097                }
098
099        }
100
101        
102        
103        @Override
104        public DoubleDBSCANClusters cluster(double[][] data) {
105                State state = new State(data.length,new NNRegionMode(data),this.noiseAsClusters);
106                return dbscan(state);
107        }
108
109        @Override
110        public DoubleDBSCANClusters cluster(DataSource<double[]> data) {
111                double[][] allData = new double[data.size()][];
112                Iterator<double[]> iterator = data.iterator();
113                for (int i = 0; i < allData.length; i++) {
114                        allData[i] = iterator.next();
115                }
116                return this.cluster(allData);
117        }
118
119        @Override
120        public int[][] performClustering(double[][] data) {
121                return cluster(data).clusters();
122        }
123
124        /**
125         * @return the epse parameter
126         */
127        public double getEps() {
128                return this.eps;
129        }
130        
131        @Override
132        public String toString() {
133                return String.format("%s: eps=%2.2f, minpts=%d, NN=%s",this.getClass().getSimpleName(),eps,minPts,this.nnf.getClass().getSimpleName());
134        }
135
136}