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
|
/*
* Copyright (C) 2012 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef KMEANS_H
#define KMEANS_H
#include <ctime>
#include <cstdlib>
#include <math.h>
// Helper functions
template <typename T, typename N>
inline void sum(T values[], int len, int dimension, int stride, N dst[]) {
int x, y;
// zero out dst vector
for (x = 0; x < dimension; x++) {
dst[x] = 0;
}
for (x = 0; x < len; x+= stride) {
for (y = 0; y < dimension; y++) {
dst[y] += values[x + y];
}
}
}
template <typename T, typename N>
inline void set(T val1[], N val2[], int dimension) {
int x;
for (x = 0; x < dimension; x++) {
val1[x] = val2[x];
}
}
template <typename T, typename N>
inline void add(T val[], N dst[], int dimension) {
int x;
for (x = 0; x < dimension; x++) {
dst[x] += val[x];
}
}
template <typename T, typename N>
inline void divide(T dst[], N divisor, int dimension) {
int x;
if (divisor == 0) {
return;
}
for (x = 0; x < dimension; x++) {
dst[x] /= divisor;
}
}
/**
* Calculates euclidean distance.
*/
template <typename T, typename N>
inline N euclideanDist(T val1[], T val2[], int dimension) {
int x;
N sum = 0;
for (x = 0; x < dimension; x++) {
N diff = (N) val1[x] - (N) val2[x];
sum += diff * diff;
}
return sqrt(sum);
}
// K-Means
/**
* Picks k random starting points from the data set.
*/
template <typename T>
void initialPickHeuristicRandom(int k, T values[], int len, int dimension, int stride, T dst[]) {
int x, z, num_vals, cntr;
num_vals = len / stride;
cntr = 0;
srand((unsigned)time(0));
unsigned int r_vals[k];
unsigned int r;
for (x = 0; x < k; x++) {
// ensure randomly chosen value is unique
int r_check = 0;
while (r_check == 0) {
r = (unsigned int) rand() % num_vals;
r_check = 1;
for (z = 0; z < x; z++) {
if (r == r_vals[z]) {
r_check = 0;
}
}
}
r_vals[x] = r;
r *= stride;
// set dst to be randomly chosen value
set<T,T>(dst + cntr, values + r, dimension);
cntr += stride;
}
}
/**
* Finds index of closet centroid to a value
*/
template <typename T, typename N>
inline int findClosest(T values[], T oldCenters[], int dimension, int stride, int pop_size) {
int best_ind = 0;
N best_len = euclideanDist <T, N>(values, oldCenters, dimension);
int y;
for (y = stride; y < pop_size; y+=stride) {
N l = euclideanDist <T, N>(values, oldCenters + y, dimension);
if (l < best_len) {
best_len = l;
best_ind = y;
}
}
return best_ind;
}
/**
* Calculates new centroids by averaging value clusters for old centroids.
*/
template <typename T, typename N>
int calculateNewCentroids(int k, T values[], int len, int dimension, int stride, T oldCenters[],
T dst[]) {
int x, pop_size;
pop_size = k * stride;
int popularities[k];
N tmp[pop_size];
//zero popularities
memset(popularities, 0, sizeof(int) * k);
// zero dst, and tmp
for (x = 0; x < pop_size; x++) {
tmp[x] = 0;
}
// put summation for each k in tmp
for (x = 0; x < len; x+=stride) {
int best = findClosest<T, N>(values + x, oldCenters, dimension, stride, pop_size);
add<T, N>(values + x, tmp + best, dimension);
popularities[best / stride]++;
}
int ret = 0;
int y;
// divide to get centroid and set dst to result
for (x = 0; x < pop_size; x+=stride) {
divide<N, int>(tmp + x, popularities[x / stride], dimension);
for (y = 0; y < dimension; y++) {
if ((dst + x)[y] != (T) ((tmp + x)[y])) {
ret = 1;
}
}
set(dst + x, tmp + x, dimension);
}
return ret;
}
/**
* Runs the k-means algorithm on dataset values with some initial centroids.
*/
template <typename T, typename N>
void runKMeans(int k, T finalCentroids[], T values[], int len, int dimension, int stride,
int iterations){
int k_len = k * stride;
int x;
T initialPicks [k_len];
initialPickHeuristicRandom<T>(k, values, len, dimension, stride, initialPicks);
// zero newCenters
for (x = 0; x < k_len; x++) {
finalCentroids[x] = 0;
}
T * c1 = initialPicks;
T * c2 = finalCentroids;
T * temp;
int ret = 1;
for (x = 0; x < iterations; x++) {
ret = calculateNewCentroids<T, N>(k, values, len, dimension, stride, c1, c2);
temp = c1;
c1 = c2;
c2 = temp;
if (ret == 0) {
x = iterations;
}
}
set<T, T>(finalCentroids, c1, dimension);
}
/**
* Sets each value in values to the closest centroid.
*/
template <typename T, typename N>
void applyCentroids(int k, T centroids[], T values[], int len, int dimension, int stride) {
int x, pop_size;
pop_size = k * stride;
for (x = 0; x < len; x+= stride) {
int best = findClosest<T, N>(values + x, centroids, dimension, stride, pop_size);
set<T, T>(values + x, centroids + best, dimension);
}
}
#endif // KMEANS_H
|
