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.list.array.TIntArrayList;
033
034import java.util.Random;
035
036import org.openimaj.experiment.evaluation.AnalysisResult;
037
038
039/**
040 * Wraps the functionality of any {@link ClusterAnalyser} as corrected by
041 * a Random baseline. This implementation follows that of cluster eval:
042 * http://chris.de-vries.id.au/2013/06/clustereval-10-release.html
043 * 
044 * @author Sina Samangooei (ss@ecs.soton.ac.uk)
045 *
046 * @param <ANNER>
047 * @param <ANNYS>
048 */
049public class RandomBaselineClusterAnalyser<
050                ANNER extends ClusterAnalyser<ANNYS>,
051                ANNYS extends RandomBaselineWrappable & AnalysisResult> 
052        implements 
053                ClusterAnalyser<RandomBaselineClusterAnalysis<ANNYS>>
054        {
055
056        private static final int NUMBER_OF_TRIALS = 100;
057        private ANNER ann;
058        private int trials;
059        private Random random;
060        /**
061         * @param analyser the underlying analyser
062         * 
063         */
064        public RandomBaselineClusterAnalyser(ANNER analyser) {
065                this.ann = analyser;
066                this.trials = NUMBER_OF_TRIALS;
067                this.random = new Random();
068        }
069        
070        /**
071         * @param analyser
072         * @param trials the number of random baselines to try, finding an average random score
073         */
074        public RandomBaselineClusterAnalyser(ANNER analyser, int trials) {
075                this.ann = analyser;
076                this.trials = trials;
077                this.random = new Random();
078        }
079        
080        /**
081         * @param analyser
082         * @param trials the number of random baselines to try, finding an average random score
083         * @param seed 
084         */
085        public RandomBaselineClusterAnalyser(ANNER analyser, int trials, long seed) {
086                this.ann = analyser;
087                this.trials = trials;
088                this.random = new Random(seed);
089        }
090        @Override
091        public RandomBaselineClusterAnalysis<ANNYS> analyse(int[][] correct,int[][] estimated) {
092                ANNYS score = ann.analyse(correct, estimated);
093                ANNYS randscore = ann.analyse(correct, baseline(estimated));
094                double meanrand = randscore.score();
095                for (int i = 0; i < this.trials; i++) {
096                        randscore = ann.analyse(correct, baseline(estimated));
097                        meanrand += randscore.score(); 
098                }
099                meanrand /= (trials+1);
100                return new RandomBaselineClusterAnalysis<ANNYS>(score,meanrand );
101        }
102        private int[][] baseline(int[][] estimated) {
103                TIntArrayList items = new TIntArrayList();
104                for (int[] is : estimated) {
105                        for (int i = 0; i < is.length; i++) {
106                                items.add(is[i]);
107                        }
108                }
109                int[][] baseline = new int[estimated.length][];
110                items.shuffle(this.random);
111                int[] itemsArr = items.toArray();
112                int seen = 0;
113                for (int i = 0; i < baseline.length; i++) {
114                        int needed = estimated[i].length;
115                        baseline[i] = new int[needed];
116                        System.arraycopy(itemsArr, seen, baseline[i], 0, needed);
117                        seen+=needed;
118                }
119                
120                return baseline;
121        }
122
123}