org.openimaj.ml.clustering.kmeans.HierarchicalDoubleKMeans Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of clustering Show documentation
Show all versions of clustering Show documentation
Various clustering algorithm implementations for all primitive types including random, random forest, K-Means (Exact, Hierarchical and Approximate), ...
/*
AUTOMATICALLY GENERATED BY jTemp FROM
/Users/jsh2/Work/openimaj/target/checkout/machine-learning/clustering/src/main/jtemp/org/openimaj/ml/clustering/kmeans/Hierarchical#T#KMeans.jtemp
*/
/**
* Copyright (c) 2011, The University of Southampton and the individual contributors.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without modification,
* are permitted provided that the following conditions are met:
*
* * Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
*
* * Redistributions in binary form must reproduce the above copyright notice,
* this list of conditions and the following disclaimer in the documentation
* and/or other materials provided with the distribution.
*
* * Neither the name of the University of Southampton nor the names of its
* contributors may be used to endorse or promote products derived from this
* software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
* ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
package org.openimaj.ml.clustering.kmeans;
import gnu.trove.list.array.TIntArrayList;
import gnu.trove.map.hash.TIntObjectHashMap;
import org.openimaj.citation.annotation.Reference;
import org.openimaj.citation.annotation.ReferenceType;
import org.openimaj.data.DataSource;
import org.openimaj.data.IndexedViewDataSource;
import org.openimaj.knn.DoubleNearestNeighbours;
import org.openimaj.ml.clustering.IndexClusters;
import org.openimaj.ml.clustering.SpatialClusterer;
import org.openimaj.ml.clustering.assignment.HardAssigner;
import org.openimaj.ml.clustering.kmeans.HierarchicalDoubleKMeansResult.Node;
import org.openimaj.util.pair.IntDoublePair;
/**
* Hierarchical Double K-Means clustering ({@link HierarchicalDoubleKMeans}) is a simple
* hierarchical version of DoubleKMeans. The algorithm recursively applies
* @{link DoubleKMeans} to create more refined partitions of the data.
*
* @author Sina Samangooei ([email protected])
* @author Jonathon Hare ([email protected])
*/
@Reference(
type = ReferenceType.Inproceedings,
author = { "David. Nist\'er", "Henrik. Stew\'enius" },
title = "Scalable Recognition with a Vocabulary Tree",
year = "2006",
booktitle = "CVPR",
pages = { "2161", "", "2168" },
customData = {
"Date-Added", "2010-11-12 09:33:18 +0000",
"Date-Modified", "2010-11-22 15:11:22 +0000"
}
)
public class HierarchicalDoubleKMeans implements SpatialClusterer {
/** data dimensionality */
int M;
/** K clusters per node */
int K;
/** KMeans configuration */
KMeansConfiguration conf;
/** Depth of the tree */
int depth;
/**
* Construct a new {@link HierarchicalDoubleKMeans} with the given parameters.
*
* @param config configuration for the underlying kmeans clustering.
* @param M Data dimensionality.
* @param K Number of clusters per node.
* @param depth Tree depth.
*/
public HierarchicalDoubleKMeans(KMeansConfiguration config, int M, int K, int depth) {
this.conf = config;
this.M = M;
this.K = K;
this.depth = depth;
}
/**
* Construct a new {@link HierarchicalDoubleKMeans} with the given parameters.
* Uses the default parameters of the {@link KMeansConfiguration}.
*
* @param M Data dimensionality.
* @param K Number of clusters per node.
* @param depth Tree depth.
*/
public HierarchicalDoubleKMeans(int M, int K, int depth) {
this(new KMeansConfiguration(), M, K, depth);
}
/**
* Extract a subset of the data to a buffer
*
* @param data Data
* @param ids Data labels
* @param id Label of data to copy
*
* @return a new buffer with a copy of the selected data.
*/
private double[][] extractSubset(final double[][] data, int[] ids, int id) {
int N = data.length;
int M = data[0].length;
int count = 0;
// count how many data points with this label there are
for (int i = 0; i < N; i++)
if (ids[i] == id)
count++;
// copy each datum to the buffer
double[][] newData = new double[count][M];
count = 0;
for (int i = 0; i < N; i++) {
if (ids[i] == id) {
System.arraycopy(data[i], 0, newData[count], 0, M);
count++;
}
}
return newData;
}
/**
* Compute HierarchicalDoubleKMeans clustering.
*
* @param data Data to cluster.
* @param K Number of clusters for this node.
* @param height Tree height.
*
* @return a new HierarchicalDoubleKMeans node representing a sub-clustering.
**/
private Node trainLevel(final double[][] data, int K, int height) {
Node node = new Node();
node.children = (height == 1) ? null : new Node[K];
DoubleKMeans kmeans = newDoubleKMeans(K);
node.result = kmeans.cluster(data);
HardAssigner assigner = node.result.defaultHardAssigner();
if (height > 1) {
int[] ids = assigner.assign(data);
for (int k = 0; k < K; k++) {
double[][] partition = extractSubset(data, ids, k);
int partitionK = Math.min(K, partition.length);
node.children[k] = trainLevel(partition, partitionK, height - 1);
}
}
return node;
}
/**
* Compute HierarchicalDoubleKMeans clustering.
*
* @param data Data to cluster.
* @param K Number of clusters for this node.
* @param height Tree height.
*
* @return a new HierarchicalDoubleKMeans node representing a sub-clustering.
**/
private Node trainLevel(final DataSource data, int K, int height) {
Node node = new Node();
node.children = (height == 1) ? null : new Node[K];
DoubleKMeans kmeans = newDoubleKMeans(K);
node.result = kmeans.cluster(data);
HardAssigner assigner = node.result.defaultHardAssigner();
if (height > 1) {
final TIntObjectHashMap assignments = new TIntObjectHashMap();
final double[][] tmp = new double[1][M];
for (int i = 0; i < data.size(); i++) {
data.getData(i, i + 1, tmp);
final int asgn = assigner.assign(tmp[0]);
TIntArrayList ids = assignments.get(asgn);
if (ids == null)
assignments.put(asgn, ids = new TIntArrayList());
ids.add(i);
}
for (int k = 0; k < K; k++) {
final int[] indexes = assignments.get(k).toArray();
final DataSource partition = new IndexedViewDataSource(data, indexes);
final int partitionK = Math.min(K, partition.size());
node.children[k] = trainLevel(partition, partitionK, height - 1);
}
}
return node;
}
@Override
public HierarchicalDoubleKMeansResult cluster(final double[][] data) {
HierarchicalDoubleKMeansResult result = new HierarchicalDoubleKMeansResult();
result.K = K;
result.M = M;
result.depth = depth;
result.root = trainLevel(data, Math.min(K, data.length), depth);
return result;
}
@Override
public int[][] performClustering(double[][] data) {
HierarchicalDoubleKMeansResult clusters = this.cluster(data);
return new IndexClusters(clusters.defaultHardAssigner().assign(data)).clusters();
}
@Override
public HierarchicalDoubleKMeansResult cluster(DataSource data) {
HierarchicalDoubleKMeansResult result = new HierarchicalDoubleKMeansResult();
result.K = K;
result.M = M;
result.depth = depth;
result.root = trainLevel(data, Math.min(K, data.size()), depth);
return result;
}
private DoubleKMeans newDoubleKMeans(int K) {
KMeansConfiguration newConf = conf.clone();
newConf.setK(K);
return new DoubleKMeans(newConf);
}
}