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
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
|
/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
/**
* Class that can run the hierarchical clustering algorithm with the given
* parameters.
*
* @param distance
* Function that should return the distance between two items.
* Defaults to clusterlib.euclidean_distance.
* @param merge
* Function that should take in two items and return a merged one.
* Defaults to clusterlib.average_linkage.
* @param threshold
* The maximum distance between two items for which their clusters
* can be merged.
*/
function HierarchicalClustering(distance, merge, threshold) {
this.distance = distance || clusterlib.euclidean_distance;
this.merge = merge || clusterlib.average_linkage;
this.threshold = threshold == undefined ? Infinity : threshold;
}
HierarchicalClustering.prototype = {
/**
* Run the hierarchical clustering algorithm on the given items to produce
* a final set of clusters. Uses the parameters set in the constructor.
*
* @param items
* An array of "things" to cluster - this is the domain-specific
* collection you're trying to cluster (colors, points, etc.)
* @param snapshotGap
* How many iterations of the clustering algorithm to wait between
* calling the snapshotCallback
* @param snapshotCallback
* If provided, will be called as clusters are merged to let you view
* the progress of the algorithm. Passed the current array of
* clusters, cached distances, and cached closest clusters.
*
* @return An array of merged clusters. The represented item can be
* found in the "item" property of the cluster.
*/
cluster: function HC_cluster(items, snapshotGap, snapshotCallback) {
// array of all remaining clusters
let clusters = [];
// 2D matrix of distances between each pair of clusters, indexed by key
let distances = [];
// closest cluster key for each cluster, indexed by key
let neighbors = [];
// an array of all clusters, but indexed by key
let clustersByKey = [];
// set up clusters from the initial items array
for (let index = 0; index < items.length; index++) {
let cluster = {
// the item this cluster represents
item: items[index],
// a unique key for this cluster, stays constant unless merged itself
key: index,
// index of cluster in clusters array, can change during any merge
index: index,
// how many clusters have been merged into this one
size: 1
};
clusters[index] = cluster;
clustersByKey[index] = cluster;
distances[index] = [];
neighbors[index] = 0;
}
// initialize distance matrix and cached neighbors
for (let i = 0; i < clusters.length; i++) {
for (let j = 0; j <= i; j++) {
var dist = (i == j) ? Infinity :
this.distance(clusters[i].item, clusters[j].item);
distances[i][j] = dist;
distances[j][i] = dist;
if (dist < distances[i][neighbors[i]]) {
neighbors[i] = j;
}
}
}
// merge the next two closest clusters until none of them are close enough
let next = null, i = 0;
for (; next = this.closestClusters(clusters, distances, neighbors); i++) {
if (snapshotCallback && (i % snapshotGap) == 0) {
snapshotCallback(clusters);
}
this.mergeClusters(clusters, distances, neighbors, clustersByKey,
clustersByKey[next[0]], clustersByKey[next[1]]);
}
return clusters;
},
/**
* Once we decide to merge two clusters in the cluster method, actually
* merge them. Alters the given state of the algorithm.
*
* @param clusters
* The array of all remaining clusters
* @param distances
* Cached distances between pairs of clusters
* @param neighbors
* Cached closest clusters
* @param clustersByKey
* Array of all clusters, indexed by key
* @param cluster1
* First cluster to merge
* @param cluster2
* Second cluster to merge
*/
mergeClusters: function HC_mergeClus(clusters, distances, neighbors,
clustersByKey, cluster1, cluster2) {
let merged = { item: this.merge(cluster1.item, cluster2.item),
left: cluster1,
right: cluster2,
key: cluster1.key,
size: cluster1.size + cluster2.size };
clusters[cluster1.index] = merged;
clusters.splice(cluster2.index, 1);
clustersByKey[cluster1.key] = merged;
// update distances with new merged cluster
for (let i = 0; i < clusters.length; i++) {
var ci = clusters[i];
var dist;
if (cluster1.key == ci.key) {
dist = Infinity;
} else if (this.merge == clusterlib.single_linkage) {
dist = distances[cluster1.key][ci.key];
if (distances[cluster1.key][ci.key] >
distances[cluster2.key][ci.key]) {
dist = distances[cluster2.key][ci.key];
}
} else if (this.merge == clusterlib.complete_linkage) {
dist = distances[cluster1.key][ci.key];
if (distances[cluster1.key][ci.key] <
distances[cluster2.key][ci.key]) {
dist = distances[cluster2.key][ci.key];
}
} else if (this.merge == clusterlib.average_linkage) {
dist = (distances[cluster1.key][ci.key] * cluster1.size
+ distances[cluster2.key][ci.key] * cluster2.size)
/ (cluster1.size + cluster2.size);
} else {
dist = this.distance(ci.item, cluster1.item);
}
distances[cluster1.key][ci.key] = distances[ci.key][cluster1.key]
= dist;
}
// update cached neighbors
for (let i = 0; i < clusters.length; i++) {
var key1 = clusters[i].key;
if (neighbors[key1] == cluster1.key ||
neighbors[key1] == cluster2.key) {
let minKey = key1;
for (let j = 0; j < clusters.length; j++) {
var key2 = clusters[j].key;
if (distances[key1][key2] < distances[key1][minKey]) {
minKey = key2;
}
}
neighbors[key1] = minKey;
}
clusters[i].index = i;
}
},
/**
* Given the current state of the algorithm, return the keys of the two
* clusters that are closest to each other so we know which ones to merge
* next.
*
* @param clusters
* The array of all remaining clusters
* @param distances
* Cached distances between pairs of clusters
* @param neighbors
* Cached closest clusters
*
* @return An array of two keys of clusters to merge, or null if there are
* no more clusters close enough to merge
*/
closestClusters: function HC_closestClus(clusters, distances, neighbors) {
let minKey = 0, minDist = Infinity;
for (let i = 0; i < clusters.length; i++) {
var key = clusters[i].key;
if (distances[key][neighbors[key]] < minDist) {
minKey = key;
minDist = distances[key][neighbors[key]];
}
}
if (minDist < this.threshold) {
return [minKey, neighbors[minKey]];
}
return null;
}
};
var clusterlib = {
hcluster: function hcluster(items, distance, merge, threshold, snapshotGap,
snapshotCallback) {
return (new HierarchicalClustering(distance, merge, threshold))
.cluster(items, snapshotGap, snapshotCallback);
},
single_linkage: function single_linkage(cluster1, cluster2) {
return cluster1;
},
complete_linkage: function complete_linkage(cluster1, cluster2) {
return cluster1;
},
average_linkage: function average_linkage(cluster1, cluster2) {
return cluster1;
},
euclidean_distance: function euclidean_distance(v1, v2) {
let total = 0;
for (let i = 0; i < v1.length; i++) {
total += Math.pow(v2[i] - v1[i], 2);
}
return Math.sqrt(total);
},
manhattan_distance: function manhattan_distance(v1, v2) {
let total = 0;
for (let i = 0; i < v1.length; i++) {
total += Math.abs(v2[i] - v1[i]);
}
return total;
},
max_distance: function max_distance(v1, v2) {
let max = 0;
for (let i = 0; i < v1.length; i++) {
max = Math.max(max, Math.abs(v2[i] - v1[i]));
}
return max;
}
};
|