1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
|
package de.lmu.ifi.dbs.elki.algorithm.clustering.kmeans.initialization;
/*
This file is part of ELKI:
Environment for Developing KDD-Applications Supported by Index-Structures
Copyright (C) 2015
Ludwig-Maximilians-Universität München
Lehr- und Forschungseinheit für Datenbanksysteme
ELKI Development Team
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
import java.util.ArrayList;
import java.util.List;
import de.lmu.ifi.dbs.elki.data.NumberVector;
import de.lmu.ifi.dbs.elki.database.Database;
import de.lmu.ifi.dbs.elki.database.datastore.DataStoreFactory;
import de.lmu.ifi.dbs.elki.database.datastore.DataStoreUtil;
import de.lmu.ifi.dbs.elki.database.datastore.WritableDoubleDataStore;
import de.lmu.ifi.dbs.elki.database.ids.ArrayModifiableDBIDs;
import de.lmu.ifi.dbs.elki.database.ids.DBIDIter;
import de.lmu.ifi.dbs.elki.database.ids.DBIDUtil;
import de.lmu.ifi.dbs.elki.database.ids.DBIDVar;
import de.lmu.ifi.dbs.elki.database.ids.DBIDs;
import de.lmu.ifi.dbs.elki.database.query.distance.DistanceQuery;
import de.lmu.ifi.dbs.elki.database.relation.Relation;
import de.lmu.ifi.dbs.elki.distance.distancefunction.NumberVectorDistanceFunction;
import de.lmu.ifi.dbs.elki.distance.distancefunction.PrimitiveDistanceFunction;
import de.lmu.ifi.dbs.elki.logging.Logging;
import de.lmu.ifi.dbs.elki.logging.progress.FiniteProgress;
import de.lmu.ifi.dbs.elki.math.MathUtil;
import de.lmu.ifi.dbs.elki.utilities.documentation.Reference;
import de.lmu.ifi.dbs.elki.utilities.exceptions.AbortException;
import de.lmu.ifi.dbs.elki.utilities.optionhandling.AbstractParameterizer;
/**
* PAM initialization for k-means (and of course, PAM).
*
* Reference:
* <p>
* Clustering my means of Medoids<br />
* Kaufman, L. and Rousseeuw, P.J.<br />
* in: Statistical Data Analysis Based on the L_1–Norm and Related Methods
* </p>
*
* @author Erich Schubert
*
* @param <O> Object type for KMedoids initialization
*/
@Reference(title = "Clustering my means of Medoids", //
authors = "Kaufman, L. and Rousseeuw, P.J.", //
booktitle = "Statistical Data Analysis Based on the L_1–Norm and Related Methods")
public class PAMInitialMeans<O> implements KMeansInitialization<NumberVector>, KMedoidsInitialization<O> {
/**
* Class logger.
*/
private static final Logging LOG = Logging.getLogger(PAMInitialMeans.class);
/**
* Constructor.
*/
public PAMInitialMeans() {
super();
}
@Override
public <T extends NumberVector, V extends NumberVector> List<V> chooseInitialMeans(Database database, Relation<T> relation, int k, NumberVectorDistanceFunction<? super T> distanceFunction, NumberVector.Factory<V> factory) {
// Ugly cast; but better than code duplication.
@SuppressWarnings("unchecked")
Relation<O> rel = (Relation<O>) relation;
// Get a distance query
@SuppressWarnings("unchecked")
final PrimitiveDistanceFunction<? super O> distF = (PrimitiveDistanceFunction<? super O>) distanceFunction;
final DistanceQuery<O> distQ = database.getDistanceQuery(rel, distF);
DBIDs medids = chooseInitialMedoids(k, rel.getDBIDs(), distQ);
List<V> medoids = new ArrayList<>(k);
for(DBIDIter iter = medids.iter(); iter.valid(); iter.advance()) {
medoids.add(factory.newNumberVector(relation.get(iter)));
}
return medoids;
}
@Override
public DBIDs chooseInitialMedoids(int k, DBIDs ids, DistanceQuery<? super O> distQ) {
ArrayModifiableDBIDs medids = DBIDUtil.newArray(k);
DBIDVar bestid = DBIDUtil.newVar();
WritableDoubleDataStore mindist = null;
// First mean is chosen by having the smallest distance sum to all others.
{
double best = Double.POSITIVE_INFINITY;
WritableDoubleDataStore newd = null;
FiniteProgress prog = LOG.isVerbose() ? new FiniteProgress("Choosing initial mean", ids.size(), LOG) : null;
for(DBIDIter iter = ids.iter(); iter.valid(); iter.advance()) {
if(newd == null) {
newd = DataStoreUtil.makeDoubleStorage(ids, DataStoreFactory.HINT_HOT | DataStoreFactory.HINT_TEMP);
}
int sum = 0;
for(DBIDIter iter2 = ids.iter(); iter2.valid(); iter2.advance()) {
double d = distQ.distance(iter, iter2);
sum += d;
newd.putDouble(iter2, d);
}
if(sum < best) {
best = sum;
bestid.set(iter);
if(mindist != null) {
mindist.destroy();
}
mindist = newd;
newd = null;
}
LOG.incrementProcessed(prog);
}
LOG.ensureCompleted(prog);
if(newd != null) {
newd.destroy();
}
medids.add(bestid);
}
assert(mindist != null);
// Subsequent means optimize the full criterion.
FiniteProgress prog = LOG.isVerbose() ? new FiniteProgress("Choosing initial centers", k, LOG) : null;
LOG.incrementProcessed(prog); // First one was just chosen.
for(int i = 1; i < k; i++) {
double best = Double.POSITIVE_INFINITY;
WritableDoubleDataStore bestd = null, newd = null;
for(DBIDIter iter = ids.iter(); iter.valid(); iter.advance()) {
if(medids.contains(iter)) {
continue;
}
if(newd == null) {
newd = DataStoreUtil.makeDoubleStorage(ids, DataStoreFactory.HINT_HOT | DataStoreFactory.HINT_TEMP);
}
double sum = 0.;
for(DBIDIter iter2 = ids.iter(); iter2.valid(); iter2.advance()) {
double v = MathUtil.min(distQ.distance(iter, iter2), mindist.doubleValue(iter2));
sum += v;
newd.put(iter2, v);
}
if(sum < best) {
best = sum;
bestid.set(iter);
if(bestd != null) {
bestd.destroy();
}
bestd = newd;
newd = null;
}
}
if(bestd == null) {
throw new AbortException("No median found that improves the criterion function?!? Too many infinite distances.");
}
medids.add(bestid);
if(newd != null) {
newd.destroy();
}
mindist.destroy();
mindist = bestd;
LOG.incrementProcessed(prog);
}
LOG.ensureCompleted(prog);
mindist.destroy();
return medids;
}
/**
* Parameterization class.
*
* @author Erich Schubert
*
* @apiviz.exclude
*/
public static class Parameterizer<V> extends AbstractParameterizer {
@Override
protected PAMInitialMeans<V> makeInstance() {
return new PAMInitialMeans<>();
}
}
}
|