/
hierarchical.swift
211 lines (194 loc) · 6.33 KB
/
hierarchical.swift
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
//
// hierarchicalClustering.swift
// deepvis
//
// Created by sdq on 8/15/17.
// Copyright © 2017 sdq. All rights reserved.
//
import Foundation
import Accelerate
class HierarchicalClustering {
static let sharedInstance = HierarchicalClustering()
fileprivate init() {}
//clustering number K
var k:Int = 2
//vectors
var vectors:[Vector] = []
//MARK: Public
//check parameters
func checkAllParameters() -> Bool {
if k < 1 { return false }
if vectors.count < k { return false }
return true
}
//add vectors
func addVector(_ newVector:Vector) {
vectors.append(newVector)
}
func addVectors(_ newVectors:Matrix) {
for newVector:Vector in newVectors {
addVector(newVector)
}
}
func reset() {
vectors.removeAll()
}
func hierarchicalClustering(completion:(_ success:Bool, _ distMat:Matrix, _ clusters:[[Vector]])->()) {
if vectors.count == 0 || vectors.count < k {
return
}
if k == 1 {
completion(true, [[]], [vectors])
}
// 1. Initialize Clusters
var clusters: [[Vector]] = []
for vector in vectors {
clusters.append([vector])
}
// 2. Distance Matrix
let m = vectors.count
let zeroVec:Vector = [Double](repeating: 0.0, count: m)
var distMat:Matrix = [Vector](repeating: zeroVec, count: m)
for i in 0..<m {
for j in 0..<m {
if i != j {
distMat[i][j] = clusterDistance(c1: clusters[i], c2: clusters[j])
}
}
}
let initDistMat = distMat
// 3. Clustering
var q = m
while q > k {
//(1) find 2 nearest clusters
var shortestDist = 10000.0
var im = 0
var jm = 1
for i in 0..<q {
for j in i..<q {
if i != j {
if distMat[i][j] < shortestDist {
shortestDist = distMat[i][j]
im = i
jm = j
}
}
}
}
//(2) merge them
clusters[im] = clusters[im] + clusters[jm]
//(3) edit clusters
clusters.remove(at: jm)
//(4) edit distMat
q -= 1
let zeroVec:Vector = [Double](repeating: 0.0, count: q)
distMat = [Vector](repeating: zeroVec, count: q)
for i in 0..<q {
for j in 0..<q {
if i != j {
distMat[i][j] = clusterDistance(c1: clusters[i], c2: clusters[j])
}
}
}
}
//return final clusters
completion(true, initDistMat, clusters)
}
func hierarchicalization(completion:(_ success:Bool, _ links:[(Int,Int)])->()) {
if vectors.count == 0 || vectors.count < k {
return
}
// 1. Initialize Clusters
var clusters: [[Vector]] = []
for vector in vectors {
clusters.append([vector])
}
let m = vectors.count
var clusterItems: [[Int]] = []
for i in 0..<m {
clusterItems.append([i])
}
// 2. Distance Matrix
let zeroVec:Vector = [Double](repeating: 0.0, count: m)
var distMat:Matrix = [Vector](repeating: zeroVec, count: m)
for i in 0..<m {
for j in 0..<m {
if i != j {
distMat[i][j] = clusterDistance(c1: clusters[i], c2: clusters[j])
}
}
}
var q = m
var merged:[Int] = []
var links:[(Int,Int)] = []
while q > 1 {
//(1) find 2 nearest clusters
var shortestDist = 10000.0
var im = 0
var jm = 1
for i in 0..<m {
if merged.contains(i) {
continue
}
for j in i..<m {
if merged.contains(j) {
continue
}
if i != j {
if distMat[i][j] < shortestDist {
shortestDist = distMat[i][j]
im = i
jm = j
}
}
}
}
//(2) merge them
let clusterim = meanVector(inputVectors: clusters[im])
let clusterjm = meanVector(inputVectors: clusters[jm])
var mindist = 10000.0
var imm = clusterItems[im][0]
var jmm = clusterItems[jm][0]
for ii in clusterItems[im] {
let newDistance = euclideanDistance(vec1: vectors[ii], vec2: clusterim)
if newDistance < mindist {
imm = ii
mindist = newDistance
}
}
mindist = 10000.0
for jj in clusterItems[jm] {
let newDistance = euclideanDistance(vec1: vectors[jj], vec2: clusterjm)
if newDistance < mindist {
jmm = jj
}
}
links.append((imm,jmm))
clusters[im] = clusters[im] + clusters[jm]
clusterItems[im] = clusterItems[im] + clusterItems[jm]
merged.append(jm)
q -= 1
let zeroVec:Vector = [Double](repeating: 0.0, count: m)
distMat = [Vector](repeating: zeroVec, count: m)
for i in 0..<m {
if merged.contains(i) {
continue
}
for j in 0..<m {
if merged.contains(j) {
continue
}
if i != j {
distMat[i][j] = clusterDistance(c1: clusters[i], c2: clusters[j])
}
}
}
completion(true, links)
}
}
func clusterDistance(c1:[Vector], c2:[Vector]) -> Double {
let c1v = meanVector(inputVectors: c1)
let c2v = meanVector(inputVectors: c2)
return euclideanDistance(vec1: c1v, vec2: c2v)
}
}