|
1
|
|
|
'''
|
|
2
|
|
|
|
|
3
|
|
|
Wrapper for CCORE library (part of this project).
|
|
4
|
|
|
|
|
5
|
|
|
Copyright (C) 2015 Andrei Novikov ([email protected])
|
|
6
|
|
|
|
|
7
|
|
|
pyclustering is free software: you can redistribute it and/or modify
|
|
8
|
|
|
it under the terms of the GNU General Public License as published by
|
|
9
|
|
|
the Free Software Foundation, either version 3 of the License, or
|
|
10
|
|
|
(at your option) any later version.
|
|
11
|
|
|
|
|
12
|
|
|
pyclustering is distributed in the hope that it will be useful,
|
|
13
|
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
14
|
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
15
|
|
|
GNU General Public License for more details.
|
|
16
|
|
|
|
|
17
|
|
|
You should have received a copy of the GNU General Public License
|
|
18
|
|
|
along with this program. If not, see <http://www.gnu.org/licenses/>.
|
|
19
|
|
|
|
|
20
|
|
|
'''
|
|
21
|
|
|
|
|
22
|
|
|
from ctypes import *;
|
|
|
|
|
|
|
23
|
|
|
|
|
24
|
|
|
from pyclustering.core.definitions import *;
|
|
|
|
|
|
|
25
|
|
|
|
|
26
|
|
|
|
|
27
|
|
|
# API that is required for interaction with DLL.
|
|
28
|
|
|
def create_pointer_data(sample):
|
|
29
|
|
|
"Allocates memory for representing input data for processing that is described by structure 'data_representation' and returns pointer this structure."
|
|
30
|
|
|
|
|
31
|
|
|
"(in) sample - dataset for processing."
|
|
|
|
|
|
|
32
|
|
|
|
|
33
|
|
|
"Returns pointer to the data for processing."
|
|
|
|
|
|
|
34
|
|
|
|
|
35
|
|
|
input_data = data_representation();
|
|
36
|
|
|
input_data.number_objects = len(sample);
|
|
37
|
|
|
input_data.dimension = len(sample[0]);
|
|
38
|
|
|
|
|
39
|
|
|
pointer_objects = (POINTER(c_double) * input_data.number_objects)();
|
|
40
|
|
|
|
|
41
|
|
|
for index in range(0, input_data.number_objects):
|
|
42
|
|
|
point = (c_double * input_data.dimension)();
|
|
43
|
|
|
for dimension in range(0, input_data.dimension):
|
|
44
|
|
|
point[dimension] = sample[index][dimension];
|
|
45
|
|
|
|
|
46
|
|
|
pointer_objects[index] = cast(point, POINTER(c_double));
|
|
47
|
|
|
|
|
48
|
|
|
input_data.pointer_objects = cast(pointer_objects, POINTER(POINTER(c_double)));
|
|
49
|
|
|
input_data = pointer(input_data);
|
|
50
|
|
|
|
|
51
|
|
|
return input_data;
|
|
52
|
|
|
|
|
53
|
|
|
|
|
54
|
|
|
def extract_clusters(ccore_result):
|
|
55
|
|
|
"Parse clustering result that is provided by the CCORE. Return Python list of clusters."
|
|
56
|
|
|
|
|
57
|
|
|
"(in) ccore_result - pointer to clustering result that has been returned by CCORE."
|
|
|
|
|
|
|
58
|
|
|
|
|
59
|
|
|
"Returns Python list of clusters."
|
|
|
|
|
|
|
60
|
|
|
|
|
61
|
|
|
pointer_clustering_result = cast(ccore_result, POINTER(clustering_result)); # clustering_result * clusters
|
|
62
|
|
|
number_clusters = pointer_clustering_result[0].number_clusters;
|
|
63
|
|
|
|
|
64
|
|
|
list_of_clusters = [];
|
|
65
|
|
|
|
|
66
|
|
|
for index_cluster in range(0, number_clusters):
|
|
67
|
|
|
clusters = cast(pointer_clustering_result[0].pointer_clusters, POINTER(cluster_representation)); # cluster_representation * cluster
|
|
68
|
|
|
|
|
69
|
|
|
objects = cast(clusters[index_cluster].pointer_objects, POINTER(c_uint)); # cluster->objects (unsigned int *)
|
|
70
|
|
|
|
|
71
|
|
|
list_of_clusters.append([]);
|
|
72
|
|
|
pointer_container = list_of_clusters[index_cluster];
|
|
73
|
|
|
|
|
74
|
|
|
for index_object in range(0, clusters[index_cluster].number_objects):
|
|
75
|
|
|
pointer_container.append(objects[index_object]);
|
|
76
|
|
|
|
|
77
|
|
|
return list_of_clusters;
|
|
78
|
|
|
|
|
79
|
|
|
|
|
80
|
|
|
def extract_dynamics(ccore_result):
|
|
81
|
|
|
"Parse dynamic result that is provided by the CCORE. Return Python tuple that represent dynamics (times, dynamic)."
|
|
82
|
|
|
|
|
83
|
|
|
"(in) ccore_result - pointer to dynamic result that has been returned by CCORE."
|
|
|
|
|
|
|
84
|
|
|
|
|
85
|
|
|
"Returns Python tuple dynamic (times, dynamic)."
|
|
|
|
|
|
|
86
|
|
|
|
|
87
|
|
|
pointer_dynamic_result = cast(ccore_result, POINTER(dynamic_result)); # dynamic_result * pointer_dynamic_result
|
|
88
|
|
|
size_dynamic = pointer_dynamic_result[0].size_dynamic;
|
|
89
|
|
|
size_network = pointer_dynamic_result[0].size_network;
|
|
90
|
|
|
|
|
91
|
|
|
pointer_time = cast(pointer_dynamic_result[0].times, POINTER(c_double));
|
|
92
|
|
|
pointer_pointer_dynamic = cast(pointer_dynamic_result[0].dynamic, POINTER(POINTER(c_double)));
|
|
93
|
|
|
|
|
94
|
|
|
times = [];
|
|
95
|
|
|
dynamic = [];
|
|
96
|
|
|
|
|
97
|
|
|
for index in range(0, size_dynamic):
|
|
98
|
|
|
times.append(pointer_time[index]);
|
|
99
|
|
|
dynamic.append([]);
|
|
100
|
|
|
|
|
101
|
|
|
pointer_dynamic = cast(pointer_pointer_dynamic[index], POINTER(c_double));
|
|
102
|
|
|
|
|
103
|
|
|
for object_dynamic in range(0, size_network):
|
|
104
|
|
|
dynamic[index].append(pointer_dynamic[object_dynamic]);
|
|
105
|
|
|
|
|
106
|
|
|
return (times, dynamic);
|
|
107
|
|
|
|
|
108
|
|
|
def extract_pyclustering_package(ccore_package_pointer):
|
|
109
|
|
|
if (ccore_package_pointer == 0):
|
|
110
|
|
|
return [];
|
|
111
|
|
|
|
|
112
|
|
|
pointer_package = cast(ccore_package_pointer, POINTER(pyclustering_package));
|
|
113
|
|
|
size = pointer_package[0].size;
|
|
114
|
|
|
type_package = pointer_package[0].type;
|
|
115
|
|
|
|
|
116
|
|
|
result = [];
|
|
117
|
|
|
pointer_data = None;
|
|
118
|
|
|
|
|
119
|
|
|
if (type_package == pyclustering_type_data.PYCLUSTERING_TYPE_INT):
|
|
120
|
|
|
pointer_data = cast(pointer_package[0].data, POINTER(c_int));
|
|
121
|
|
|
|
|
122
|
|
|
elif (type_package == pyclustering_type_data.PYCLUSTERING_TYPE_UNSIGNED_INT):
|
|
123
|
|
|
pointer_data = cast(pointer_package[0].data, POINTER(c_uint));
|
|
124
|
|
|
|
|
125
|
|
|
elif (type_package == pyclustering_type_data.PYCLUSTERING_TYPE_FLOAT):
|
|
126
|
|
|
pointer_data = cast(pointer_package[0].data, POINTER(c_float));
|
|
127
|
|
|
|
|
128
|
|
|
elif (type_package == pyclustering_type_data.PYCLUSTERING_TYPE_DOUBLE):
|
|
129
|
|
|
pointer_data = cast(pointer_package[0].data, POINTER(c_double));
|
|
130
|
|
|
|
|
131
|
|
|
elif (type_package == pyclustering_type_data.PYCLUSTERING_TYPE_LONG):
|
|
132
|
|
|
pointer_data = cast(pointer_package[0].data, POINTER(c_long));
|
|
133
|
|
|
|
|
134
|
|
|
elif (type_package == pyclustering_type_data.PYCLUSTERING_TYPE_UNSIGNED_LONG):
|
|
135
|
|
|
pointer_data = cast(pointer_package[0].data, POINTER(c_ulong));
|
|
136
|
|
|
|
|
137
|
|
|
elif (type_package == pyclustering_type_data.PYCLUSTERING_TYPE_SIZE_T):
|
|
138
|
|
|
pointer_data = cast(pointer_package[0].data, POINTER(c_size_t));
|
|
139
|
|
|
|
|
140
|
|
|
elif (type_package == pyclustering_type_data.PYCLUSTERING_TYPE_LIST):
|
|
141
|
|
|
# pointer_package[0].data == pyclustering_package **
|
|
142
|
|
|
pointer_data = cast(pointer_package[0].data, POINTER(POINTER(pyclustering_package)));
|
|
143
|
|
|
|
|
144
|
|
|
for index in range(0, size):
|
|
145
|
|
|
pointer_package = cast(pointer_data[index], (POINTER(pyclustering_package)));
|
|
146
|
|
|
result.append(extract_pyclustering_package(pointer_package));
|
|
147
|
|
|
|
|
148
|
|
|
else:
|
|
149
|
|
|
assert(0);
|
|
150
|
|
|
|
|
151
|
|
|
if (type_package != pyclustering_type_data.PYCLUSTERING_TYPE_LIST):
|
|
152
|
|
|
for index in range(0, size):
|
|
153
|
|
|
result.append(pointer_data[index]);
|
|
154
|
|
|
|
|
155
|
|
|
return result;
|
|
156
|
|
|
|
|
157
|
|
|
|
|
158
|
|
|
# Implemented algorithms.
|
|
159
|
|
|
def dbscan(sample, eps, min_neighbors, return_noise = False):
|
|
160
|
|
|
"Clustering algorithm DBSCAN returns allocated clusters and noise that are consisted from input data. Calculation is performed via CCORE."
|
|
161
|
|
|
|
|
162
|
|
|
"(in) data - input data that is presented as list of points (objects), each point should be represented by list or tuple."
|
|
|
|
|
|
|
163
|
|
|
"(in) eps - connectivity radius between points, points may be connected if distance between them less then the radius."
|
|
|
|
|
|
|
164
|
|
|
"(in) min_neighbors - minimum number of shared neighbors that is requied for establish links between points."
|
|
|
|
|
|
|
165
|
|
|
"(in) return_noise - if True than list of points that have been marked as noise will be returned."
|
|
|
|
|
|
|
166
|
|
|
|
|
167
|
|
|
"If return_noise is False: Returns list of allocated clusters, each cluster contains indexes of objects in list of data."
|
|
|
|
|
|
|
168
|
|
|
"If return_noise is True: Returns tuple of list of allicated clusters and list of points that are marked as noise."
|
|
|
|
|
|
|
169
|
|
|
|
|
170
|
|
|
pointer_data = create_pointer_data(sample);
|
|
171
|
|
|
|
|
172
|
|
|
ccore = cdll.LoadLibrary(PATH_DLL_CCORE_64);
|
|
173
|
|
|
result = ccore.dbscan_algorithm(pointer_data, c_double(eps), c_uint(min_neighbors));
|
|
174
|
|
|
|
|
175
|
|
|
list_of_clusters = extract_clusters(result);
|
|
176
|
|
|
ccore.free_clustering_result(result);
|
|
177
|
|
|
|
|
178
|
|
|
noise = list_of_clusters[len(list_of_clusters) - 1];
|
|
179
|
|
|
list_of_clusters.remove(noise);
|
|
180
|
|
|
|
|
181
|
|
|
if (return_noise is True):
|
|
182
|
|
|
return (list_of_clusters, noise);
|
|
183
|
|
|
else:
|
|
184
|
|
|
return list_of_clusters;
|
|
185
|
|
|
|
|
186
|
|
|
def cure(sample, number_clusters, number_represent_points, compression):
|
|
187
|
|
|
pointer_data = create_pointer_data(sample);
|
|
188
|
|
|
|
|
189
|
|
|
ccore = cdll.LoadLibrary(PATH_DLL_CCORE_64);
|
|
190
|
|
|
result = ccore.cure_algorithm(pointer_data, c_uint(number_clusters), c_uint(number_represent_points), c_double(compression));
|
|
191
|
|
|
|
|
192
|
|
|
list_of_clusters = extract_clusters(result);
|
|
193
|
|
|
|
|
194
|
|
|
ccore.free_clustering_result(result);
|
|
195
|
|
|
return list_of_clusters;
|
|
196
|
|
|
|
|
197
|
|
|
|
|
198
|
|
|
def kmeans(sample, centers, tolerance):
|
|
199
|
|
|
"Clustering algorithm K-Means returns allocated clusters. Calculation is performed via CCORE."
|
|
200
|
|
|
|
|
201
|
|
|
"(in) data - input data that is presented as list of points (objects), each point should be represented by list or tuple."
|
|
|
|
|
|
|
202
|
|
|
"(in) centers - initial coordinates of centers of clusters that are represented by list: [center1, center2, ...]."
|
|
|
|
|
|
|
203
|
|
|
"(in) tolerance - stop condition: if maximum value of change of centers of clusters is less than tolerance than algorithm will stop processing."
|
|
|
|
|
|
|
204
|
|
|
|
|
205
|
|
|
"Returns list of allocated clusters, each cluster contains indexes of objects in list of data."
|
|
|
|
|
|
|
206
|
|
|
|
|
207
|
|
|
pointer_data = create_pointer_data(sample);
|
|
208
|
|
|
pointer_centers = create_pointer_data(centers);
|
|
209
|
|
|
|
|
210
|
|
|
ccore = cdll.LoadLibrary(PATH_DLL_CCORE_64);
|
|
211
|
|
|
result = ccore.kmeans_algorithm(pointer_data, pointer_centers, c_double(tolerance));
|
|
212
|
|
|
|
|
213
|
|
|
list_of_clusters = extract_clusters(result);
|
|
214
|
|
|
|
|
215
|
|
|
ccore.free_clustering_result(result);
|
|
216
|
|
|
return list_of_clusters;
|
|
217
|
|
|
|
|
218
|
|
|
|
|
219
|
|
|
def kmedians(sample, centers, tolerance):
|
|
220
|
|
|
pointer_data = create_pointer_data(sample);
|
|
221
|
|
|
pointer_centers = create_pointer_data(centers);
|
|
222
|
|
|
|
|
223
|
|
|
ccore = cdll.LoadLibrary(PATH_DLL_CCORE_64);
|
|
224
|
|
|
result = ccore.kmedians_algorithm(pointer_data, pointer_centers, c_double(tolerance));
|
|
225
|
|
|
|
|
226
|
|
|
list_of_clusters = extract_clusters(result);
|
|
227
|
|
|
ccore.free_clustering_result(result);
|
|
228
|
|
|
|
|
229
|
|
|
return list_of_clusters;
|
|
230
|
|
|
|
|
231
|
|
|
|
|
232
|
|
|
def rock(sample, eps, number_clusters, threshold):
|
|
233
|
|
|
"Clustering algorithm ROCK returns allocated clusters and noise that are consisted from input data. Calculation is performed via CCORE."
|
|
234
|
|
|
|
|
235
|
|
|
"(in) data - input data - list of points where each point is represented by list of coordinates."
|
|
|
|
|
|
|
236
|
|
|
"(in) eps - connectivity radius (similarity threshold), points are neighbors if distance between them is less than connectivity radius."
|
|
|
|
|
|
|
237
|
|
|
"(in) number_clusters - defines number of clusters that should be allocated from the input data set."
|
|
|
|
|
|
|
238
|
|
|
"(in) threshold - value that defines degree of normalization that influences on choice of clusters for merging during processing."
|
|
|
|
|
|
|
239
|
|
|
|
|
240
|
|
|
"Returns list of allocated clusters, each cluster contains indexes of objects in list of data."
|
|
|
|
|
|
|
241
|
|
|
|
|
242
|
|
|
pointer_data = create_pointer_data(sample);
|
|
243
|
|
|
|
|
244
|
|
|
ccore = cdll.LoadLibrary(PATH_DLL_CCORE_64);
|
|
245
|
|
|
result = ccore.rock_algorithm(pointer_data, c_double(eps), c_uint(number_clusters), c_double(threshold));
|
|
246
|
|
|
|
|
247
|
|
|
list_of_clusters = extract_clusters(result);
|
|
248
|
|
|
ccore.free_clustering_result(result);
|
|
249
|
|
|
|
|
250
|
|
|
return list_of_clusters;
|
|
251
|
|
|
|
|
252
|
|
|
|
|
253
|
|
|
def xmeans(sample, centers, kmax, tolerance):
|
|
254
|
|
|
"Clustering algorithm X-Means returns allocated clusters. Calculation is performed via CCORE."
|
|
255
|
|
|
|
|
256
|
|
|
"(in) data - input data that is presented as list of points (objects), each point should be represented by list or tuple."
|
|
|
|
|
|
|
257
|
|
|
"(in) centers - initial coordinates of centers of clusters that are represented by list: [center1, center2, ...]."
|
|
|
|
|
|
|
258
|
|
|
"(in) kmax - maximum number of clusters that can be allocated."
|
|
|
|
|
|
|
259
|
|
|
|
|
260
|
|
|
"Returns list of allocated clusters, each cluster contains indexes of objects in list of data."
|
|
|
|
|
|
|
261
|
|
|
|
|
262
|
|
|
pointer_data = create_pointer_data(sample);
|
|
263
|
|
|
pointer_centers = create_pointer_data(centers);
|
|
264
|
|
|
|
|
265
|
|
|
ccore = cdll.LoadLibrary(PATH_DLL_CCORE_64);
|
|
266
|
|
|
result = ccore.xmeans_algorithm(pointer_data, pointer_centers, c_uint(kmax), c_double(tolerance));
|
|
267
|
|
|
|
|
268
|
|
|
list_of_clusters = extract_clusters(result);
|
|
269
|
|
|
|
|
270
|
|
|
ccore.free_clustering_result(result);
|
|
271
|
|
|
return list_of_clusters;
|
|
272
|
|
|
|
|
273
|
|
|
|
|
274
|
|
|
"CCORE Interface for HSYNCNET oscillatory network"
|
|
|
|
|
|
|
275
|
|
|
|
|
276
|
|
|
def hsyncnet_create_network(sample, number_clusters, initial_phases, initial_neighbors, increase_persent):
|
|
277
|
|
|
pointer_data = create_pointer_data(sample);
|
|
278
|
|
|
|
|
279
|
|
|
ccore = cdll.LoadLibrary(PATH_DLL_CCORE_64);
|
|
280
|
|
|
pointer_network = ccore.hsyncnet_create_network(pointer_data, c_uint(number_clusters), c_uint(initial_phases), c_uint(initial_neighbors), c_double(increase_persent));
|
|
281
|
|
|
|
|
282
|
|
|
return pointer_network;
|
|
283
|
|
|
|
|
284
|
|
|
|
|
285
|
|
|
def hsyncnet_destroy_network(pointer_network):
|
|
286
|
|
|
ccore = cdll.LoadLibrary(PATH_DLL_CCORE_64);
|
|
287
|
|
|
ccore.hsyncnet_destroy_network(pointer_network);
|
|
288
|
|
|
|
|
289
|
|
|
|
|
290
|
|
|
def hsyncnet_process(network_pointer, order, solution, collect_dynamic):
|
|
291
|
|
|
ccore = cdll.LoadLibrary(PATH_DLL_CCORE_64);
|
|
292
|
|
|
return ccore.hsyncnet_process(network_pointer, c_double(order), c_uint(solution), c_bool(collect_dynamic));
|
|
293
|
|
|
|
|
294
|
|
|
|
|
295
|
|
|
def hsyncnet_analyser_destroy(pointer_analyser):
|
|
296
|
|
|
ccore = cdll.LoadLibrary(PATH_DLL_CCORE_64);
|
|
297
|
|
|
ccore.syncnet_analyser_destroy(pointer_analyser);
|
|
298
|
|
|
|
|
299
|
|
|
|
annoviko /
pyclustering
Loading history...