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.examples.ml.clustering.kmeans; 031 032import java.util.Arrays; 033 034import org.openimaj.data.RandomData; 035import org.openimaj.ml.clustering.ByteCentroidsResult; 036import org.openimaj.ml.clustering.assignment.HardAssigner; 037import org.openimaj.ml.clustering.kmeans.ByteKMeans; 038 039/** 040 * Example showing how to use the Exact or Approximate KMeans clustering 041 * algorithm with in-memory data. The example sets up the clustering algorithm, 042 * generates some uniform random data to cluster, and runs the clustering 043 * algorithm. The resultant clusters are printed, and the example also shows how 044 * to perform cluster assignment (i.e. vector quantisation) in which vectors are 045 * assigned to their closest cluster. 046 * 047 * @author Jonathon Hare (jsh2@ecs.soton.ac.uk) 048 */ 049public class KMeansExample { 050 /** 051 * Main method for the example. 052 * 053 * @param args 054 * Ignored. 055 */ 056 public static void main(String[] args) { 057 // Set up the variables needed to define the clustering operation 058 final int dimensionality = 2; 059 final int numItems = 10000; 060 final int K = 4; 061 062 // Create the clusterer; there are specific types for all kinds of data 063 // (we're using byte data here). The clusters provide a couple of 064 // static methods to quickly create an exact or approximate (using a 065 // KD-Tree ensemble) K-Means algorithm. Alternatively, the K-Means 066 // clusterer can be constructed with a KMeansConfiguration object which 067 // offers fine control over the k-means algorithm, including control 068 // over the nearest-neighbour implementation (i.e. approximate or not; 069 // different distance functions, etc). 070 final ByteKMeans kmeans = ByteKMeans.createExact(dimensionality, K); 071 072 // Settings for things like the number of iterations can be set through 073 // the KMeansConfiguration either at construction time, or afterwards: 074 kmeans.getConfiguration().setMaxIterations(100); 075 076 // Generate some random data to cluster 077 final byte[][] data = RandomData.getRandomByteArray(numItems, dimensionality, Byte.MIN_VALUE, Byte.MAX_VALUE); 078 079 // Perform the clustering 080 final ByteCentroidsResult result = kmeans.cluster(data); 081 082 // Get an assigner to assign vectors to the closest cluster. You could 083 // also construct an assigner implementation manually if you want 084 // more control over the assigner. 085 final HardAssigner<byte[], ?, ?> assigner = result.defaultHardAssigner(); 086 087 // Now investigate which cluster each original data item belonged to: 088 for (int i = 0; i < 10; i++) { 089 final byte[] vector = data[i]; 090 091 // Each leaf-node of the hierarchy is assigned a unique index number 092 // between 0 and the total number of leaf-nodes. 093 final int globalClusterNumber = assigner.assign(vector); 094 095 System.out.format("%s was assigned to cluster %d\n", Arrays.toString(vector), 096 globalClusterNumber); 097 } 098 } 099}