-
Notifications
You must be signed in to change notification settings - Fork 1
/
mykmeanssp.c
291 lines (237 loc) · 7.68 KB
/
mykmeanssp.c
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
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
#define PY_SSIZE_T_CLEAN
#include <Python.h>
#include <stdio.h>
#include <stdlib.h>
#include <math.h>
const double EPS = 0.001;
const int ITER = 200;
const char* GENERAL_ERROR = "An Error Has Occurred\n";
Py_ssize_t dim = 0;
Py_ssize_t line_num = 0;
/* ============ Array Tools ============ */
/* frees 2d arrays. Array needs to be cast into void** */
void free_2d(void** mat, int rows){
int i;
for (i = 0; i < rows; i++){
free(mat[i]);
}
free(mat);
}
void Py_free_2d(void **mat, Py_ssize_t rows){
int i;
for (i = 0; i < rows; i++)
{
free(mat[i]);
}
free(mat);
}
/* Checks that an allocated pointer actually has somewhere to point to */
void pointer_check(void* ptr, const char* error_msg){
if (ptr == NULL) {
printf("%s", error_msg);
exit(1);
}
return;
}
/* Creates an array of points that are all 0.
* len - length of the array
* Checks new pointers */
double** empty_points_arr(int len){
int i, j;
double **arr = (double **)malloc(len * sizeof(double *));
pointer_check((void *)arr, GENERAL_ERROR);
for (i = 0; i < len; i++){
arr[i] = (double *)malloc(dim * sizeof(double));
pointer_check((void *)arr[i], GENERAL_ERROR);
for (j = 0; j < dim; j++){
arr[i][j] = 0.0;
}
}
return arr;
}
/* Copies the first k points from one array to a new one.
* Checks new pointers */
double** point_array_copy(double** points, int k){
int i, j;
double **new_points = (double **)malloc(k * sizeof(double *));
pointer_check((void *)new_points, GENERAL_ERROR);
for (i = 0; i < k; i++){
new_points[i] = (double *)malloc(dim * sizeof(double));
pointer_check((void *)new_points[i], GENERAL_ERROR);
/* copy the values directly and not by pointing */
for (j = 0; j < dim; j++){
new_points[i][j] = points[i][j];
}
}
return new_points;
}
/* Euclidian distance between p and q.
* Formula d=sqrt(pow(p[0]-q[0], 2) + ... + pow(p[n]-q[n], 2)) */
double distance(double* p, double* q){
int i;
double dist = 0.0;
for (i = 0; i < dim; i++){
dist += pow((p[i]-q[i]), 2);
}
return sqrt(dist);
}
/* Calculate k-means of a 2d array representing poinds */
double** kmeans(double** centroids,double** points, int k, int iter, double eps){
int i,j,l;
for (i = 0; i < iter; i++){
double **new_cents = empty_points_arr(k);
/* members saves what center assigned to each point.
* point[i] is assigned to centroids[members[i]] */
int *members = (int *)calloc(line_num, sizeof(int));
/* mem_cnt counts the number of points assigned to a center */
int *mem_cnt = (int *)calloc(k, sizeof(int));
int converg_cnt = 0;
/* foreach point*/
for (j = 0; j < line_num; j++){
int center_idx = 0;
double min_dist = -1;
/* find closest cluster*/
for (l = 0; l < k; l++){
double dist = distance(points[j], centroids[l]);
if ((min_dist == -1)||(min_dist > dist)){
center_idx = l;
min_dist = dist;
}
}
members[j] = center_idx; /* point j is member of center_idx*/
mem_cnt[center_idx] += 1; /* count member*/
}
/*step 4*/
for (j = 0; j < line_num; j++){
/* foreach point find its assigned index */
int center_idx = members[j];
for (l = 0; l < dim; l++){
/* sum the point components normalized by member count (avg point) */
new_cents[center_idx][l] += points[j][l]/mem_cnt[center_idx];
}
}
/*step 5*/
for (j = 0; j < k; j++){
/* count unchanged distances */
if (distance(new_cents[j], centroids[j]) < eps){
converg_cnt++;
}
/* also, we won't need the old centers anymore */
free(centroids[j]);
}
/* free all helping memory */
free(centroids);
free(members);
free(mem_cnt);
/* save new centers */
centroids = new_cents;
if (converg_cnt == k){
/* convergence */
break;
}
}
return centroids;
}
double** fit(double** centroids, double** points, int k, int iter, double eps) {
int i, j;
centroids = kmeans(centroids, points, k, iter, eps);
for (i = 0; i < k; i++) {
for (j = 0; j < dim-1; j++) {
printf("%.4f,",centroids[i][j]);
}
printf("%.4f\n",centroids[i][dim-1]);
}
return centroids;
}
/* Function to convert Python object to double** */
double** convert_to_double_array(PyObject* obj) {
Py_ssize_t i, j;
Py_ssize_t rows = PyList_Size(obj);
double** arr = (double**)malloc(rows * sizeof(double*));
for (i = 0; i < rows; i++) {
PyObject* row = PyList_GetItem(obj, i);
Py_ssize_t cols = PyList_Size(row);
arr[i] = (double*)malloc(cols * sizeof(double));
for (j = 0; j < cols; j++) {
PyObject* item = PyList_GetItem(row, j);
if (!PyFloat_Check(item)) {
/* Clean up and return NULL if item is not a float */
for (int k = 0; k < i; k++) {
free(arr[k]);
}
free(arr);
return NULL;
}
arr[i][j] = PyFloat_AsDouble(item);
}
}
return arr;
}
/* Function to convert double** to Python object */
PyObject* convert_to_python_object(double** arr, Py_ssize_t rows, Py_ssize_t cols) {
Py_ssize_t i, j;
PyObject* obj = PyList_New(rows);
for (i = 0; i < rows; i++) {
PyObject* row = PyList_New(cols);
for (j = 0; j < cols; j++) {
PyObject* item = PyFloat_FromDouble(arr[i][j]);
if (!item) {
/* Clean up and return NULL if conversion fails */
Py_DECREF(row);
Py_DECREF(obj);
return NULL;
}
PyList_SetItem(row, j, item);
}
PyList_SetItem(obj, i, row);
}
return obj;
}
static PyObject* fit_wrapper(PyObject *self, PyObject *args) {
PyObject* centroids_obj;
PyObject* points_obj;
int k, iter;
double eps;
/* This parses the Python arguments into Python objects */
if (!PyArg_ParseTuple(args, "OOiid", ¢roids_obj, &points_obj, &k, &iter, &eps)) {
return NULL;
}
/* Assign dimension and number of lines global variables */
line_num = PyList_Size(points_obj);
dim = line_num > 0? PyList_Size(PyList_GetItem(points_obj, 0)) : 0;
/* Convert Python objects to double arrays */
double** centroids = convert_to_double_array(centroids_obj);
double** points = convert_to_double_array(points_obj);
/* Call the fit function */
double** result = fit(centroids, points, k, iter, eps);
/* Convert the result back to a Python object */
PyObject* result_obj = convert_to_python_object(result, k, dim);
/* Free the allocated memory */
Py_free_2d((void**)points, line_num);
free_2d((void**)result, k);
return result_obj;
}
static PyMethodDef kmeansMethods[] = {
{
"fit",
fit_wrapper,
METH_VARARGS,
PyDoc_STR("Implementation of k-means clustering algorithm")
},
{NULL, NULL, 0, NULL}
};
static struct PyModuleDef mykmeanssp = {
PyModuleDef_HEAD_INIT,
"mykmeanssp",
"kmeans module",
-1,
kmeansMethods
};
PyMODINIT_FUNC PyInit_mykmeanssp(void){
PyObject *m;
m = PyModule_Create(&mykmeanssp);
if (!m) {
return NULL;
}
return m;
}