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.experiment.evaluation.cluster.analyser;
031
032import gnu.trove.map.hash.TIntLongHashMap;
033import gnu.trove.procedure.TIntLongProcedure;
034import gov.sandia.cognition.math.MathUtil;
035
036import java.util.HashSet;
037import java.util.Map;
038import java.util.Set;
039
040import org.apache.log4j.Logger;
041
042/**
043 * Create an {@link AdjustedRandomIndexAnalysis} instance 
044 * @author Sina Samangooei (ss@ecs.soton.ac.uk)
045 */
046public class AdjustedRandomIndexClusterAnalyser implements ClusterAnalyser<AdjustedRandomIndexAnalysis>{
047        final static Logger logger = Logger.getLogger(AdjustedRandomIndexAnalysis.class);
048        @Override
049        public AdjustedRandomIndexAnalysis analyse(int[][] correct, int[][] estimated) {
050                TIntLongHashMap nij = new TIntLongHashMap();
051                TIntLongHashMap ni = new TIntLongHashMap();
052                TIntLongHashMap nj = new TIntLongHashMap();
053                
054                Map<Integer,Integer> invCor = ClusterAnalyserUtils.invert(correct);
055                logger.debug("Correct keys: " + invCor.size());
056                Map<Integer,Integer> invEst = ClusterAnalyserUtils.invert(estimated);
057                logger.debug("Estimated keys: " + invCor.size());
058                Set<Integer> sharedKeys = new HashSet<Integer>();
059                sharedKeys.addAll(invCor.keySet());
060                sharedKeys.retainAll(invEst.keySet());
061                logger.debug("Shared keys: " + sharedKeys.size());
062                for (Integer index : sharedKeys) {
063                        int i = invCor.get(index);
064                        int j = invEst.get(index);
065                        nij.adjustOrPutValue(i * correct.length + j, 1, 1);
066                        ni.adjustOrPutValue(i, 1, 1);
067                        nj.adjustOrPutValue(j, 1, 1);
068                }
069                
070                final long[] sumnij = new  long[1];
071                final long[] sumni  = new long[1];
072                final long[] sumnj  = new long[1];
073                final long[] sumn   = new long[1];
074                
075                nj.forEachEntry(new TIntLongProcedure() {
076                        
077                        @Override
078                        public boolean execute(int a, long b) {
079                                if(b > 1){
080                                        sumnj[0] += MathUtil.binomialCoefficient((int)b, 2);
081                                }
082                                return true;
083                        }
084                });
085                
086                ni.forEachEntry(new TIntLongProcedure() {
087                        
088                        @Override
089                        public boolean execute(int a, long b) {
090                                if(b > 1){
091                                        sumni[0] += MathUtil.binomialCoefficient((int)b, 2);
092                                }
093                                sumn[0] += b;
094                                return true;
095                        }
096                });
097                nij.forEachEntry(new TIntLongProcedure() {
098                        
099                        @Override
100                        public boolean execute(int a, long b) {
101                                if(b > 1){
102                                        sumnij[0] += MathUtil.binomialCoefficient((int)b, 2);
103                                }
104                                return true;
105                        }
106                });
107                double bisumn = MathUtil.binomialCoefficient((int) sumn[0], 2);
108                double div = (sumni[0] * sumnj[0]) / bisumn;
109                AdjustedRandomIndexAnalysis ret = new AdjustedRandomIndexAnalysis();
110                ret.adjRandInd = (sumnij[0] - div) / (0.5 * (sumni[0] + sumnj[0]) - div);
111                
112                return ret;
113        }
114
115}