|
1
|
|
|
"""!
|
|
2
|
|
|
|
|
3
|
|
|
@brief Neural Network: Self-Organized Feature Map
|
|
4
|
|
|
@details Based on article description:
|
|
5
|
|
|
- T.Kohonen. The Self-Organizing Map. 1990.
|
|
6
|
|
|
- T.Kohonen, E.Oja, O.Simula, A.Visa, J.Kangas. Engineering Applications of the Self-Organizing Map. 1996.
|
|
7
|
|
|
|
|
8
|
|
|
@authors Andrei Novikov ([email protected])
|
|
9
|
|
|
@date 2014-2016
|
|
10
|
|
|
@copyright GNU Public License
|
|
11
|
|
|
|
|
12
|
|
|
@cond GNU_PUBLIC_LICENSE
|
|
13
|
|
|
PyClustering is free software: you can redistribute it and/or modify
|
|
14
|
|
|
it under the terms of the GNU General Public License as published by
|
|
15
|
|
|
the Free Software Foundation, either version 3 of the License, or
|
|
16
|
|
|
(at your option) any later version.
|
|
17
|
|
|
|
|
18
|
|
|
PyClustering is distributed in the hope that it will be useful,
|
|
19
|
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
20
|
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
21
|
|
|
GNU General Public License for more details.
|
|
22
|
|
|
|
|
23
|
|
|
You should have received a copy of the GNU General Public License
|
|
24
|
|
|
along with this program. If not, see <http://www.gnu.org/licenses/>.
|
|
25
|
|
|
@endcond
|
|
26
|
|
|
|
|
27
|
|
|
"""
|
|
28
|
|
|
|
|
29
|
|
|
|
|
30
|
|
|
import math;
|
|
31
|
|
|
import random;
|
|
32
|
|
|
|
|
33
|
|
|
import matplotlib.pyplot as plt;
|
|
|
|
|
|
|
34
|
|
|
|
|
35
|
|
|
import pyclustering.core.som_wrapper as wrapper;
|
|
36
|
|
|
|
|
37
|
|
|
from pyclustering.utils import euclidean_distance_sqrt;
|
|
38
|
|
|
|
|
39
|
|
|
from enum import IntEnum;
|
|
40
|
|
|
|
|
41
|
|
|
|
|
42
|
|
|
class type_conn(IntEnum):
|
|
43
|
|
|
"""!
|
|
44
|
|
|
@brief Enumeration of connection types for SOM.
|
|
45
|
|
|
|
|
46
|
|
|
@see som
|
|
47
|
|
|
|
|
48
|
|
|
"""
|
|
49
|
|
|
|
|
50
|
|
|
## Grid type of connections when each oscillator has connections with left, upper, right, lower neighbors.
|
|
51
|
|
|
grid_four = 0;
|
|
52
|
|
|
|
|
53
|
|
|
## Grid type of connections when each oscillator has connections with left, upper-left, upper, upper-right, right, right-lower, lower, lower-left neighbors.
|
|
54
|
|
|
grid_eight = 1;
|
|
55
|
|
|
|
|
56
|
|
|
## Grid type of connections when each oscillator has connections with left, upper-left, upper-right, right, right-lower, lower-left neighbors.
|
|
57
|
|
|
honeycomb = 2;
|
|
58
|
|
|
|
|
59
|
|
|
## Grid type of connections when existance of each connection is defined by the SOM rule on each step of simulation.
|
|
60
|
|
|
func_neighbor = 3;
|
|
61
|
|
|
|
|
62
|
|
|
|
|
63
|
|
|
class type_init(IntEnum):
|
|
64
|
|
|
"""!
|
|
65
|
|
|
@brief Enumeration of initialization types for SOM
|
|
66
|
|
|
|
|
67
|
|
|
@see som
|
|
68
|
|
|
|
|
69
|
|
|
"""
|
|
70
|
|
|
|
|
71
|
|
|
## Weights are randomly distributed using Gaussian distribution (0, 1).
|
|
72
|
|
|
random = 0;
|
|
73
|
|
|
|
|
74
|
|
|
## Weights are randomly distributed using Gaussian distribution (input data centroid, 1).
|
|
75
|
|
|
random_centroid = 1;
|
|
76
|
|
|
|
|
77
|
|
|
## Weights are randomly distrbiuted using Gaussian distribution (input data centroid, surface of input data).
|
|
78
|
|
|
random_surface = 2;
|
|
79
|
|
|
|
|
80
|
|
|
## Weights are distributed as a uniform grid that covers whole surface of the input data.
|
|
81
|
|
|
uniform_grid = 3;
|
|
82
|
|
|
|
|
83
|
|
|
|
|
84
|
|
|
class som_parameters:
|
|
85
|
|
|
"""!
|
|
86
|
|
|
@brief Represents SOM parameters.
|
|
87
|
|
|
|
|
88
|
|
|
"""
|
|
89
|
|
|
|
|
90
|
|
|
## Type of initialization of initial neuron weights (random, random in center of the input data, random distributed in data, ditributed in line with uniform grid).
|
|
91
|
|
|
init_type = type_init.uniform_grid;
|
|
92
|
|
|
|
|
93
|
|
|
## Initial radius (if not specified then will be calculated by SOM).
|
|
94
|
|
|
init_radius = None;
|
|
95
|
|
|
|
|
96
|
|
|
## Rate of learning.
|
|
97
|
|
|
init_learn_rate = 0.1;
|
|
98
|
|
|
|
|
99
|
|
|
## Condition when learining process should be stoped. It's used when autostop mode is used.
|
|
100
|
|
|
adaptation_threshold = 0.001;
|
|
101
|
|
|
|
|
102
|
|
|
|
|
103
|
|
|
class som:
|
|
104
|
|
|
"""!
|
|
105
|
|
|
@brief Represents self-organized feature map (SOM).
|
|
106
|
|
|
|
|
107
|
|
|
Example:
|
|
108
|
|
|
@code
|
|
109
|
|
|
# sample for training
|
|
110
|
|
|
sample_train = read_sample(file_train_sample);
|
|
111
|
|
|
|
|
112
|
|
|
# create self-organized feature map with size 5x5
|
|
113
|
|
|
network = som(5, 5, sample_train, 100);
|
|
114
|
|
|
|
|
115
|
|
|
# train network
|
|
116
|
|
|
network.train();
|
|
117
|
|
|
|
|
118
|
|
|
# simulate using another sample
|
|
119
|
|
|
sample = read_sample(file_sample);
|
|
120
|
|
|
index_winner = network.simulate(sample);
|
|
121
|
|
|
|
|
122
|
|
|
# check what it is (what it looks like?)
|
|
123
|
|
|
index_similar_objects = network.capture_objects[index_winner];
|
|
124
|
|
|
|
|
125
|
|
|
@endcode
|
|
126
|
|
|
|
|
127
|
|
|
"""
|
|
128
|
|
|
|
|
129
|
|
|
# describe network
|
|
130
|
|
|
_rows = 0;
|
|
131
|
|
|
_cols = 0;
|
|
132
|
|
|
_size = 0;
|
|
133
|
|
|
_weights = None; # Weights of each neuron (coordinates in data dimension in other words).
|
|
134
|
|
|
_award = None; # Lists of indexes of won points for each neuron.
|
|
135
|
|
|
_data = None; # Analyzed data.
|
|
136
|
|
|
_conn_type = None; # Type of connections between neuron.
|
|
137
|
|
|
|
|
138
|
|
|
# just for convenience (avoid excess calculation during learning)
|
|
139
|
|
|
_location = None; # Location in grid.
|
|
140
|
|
|
_sqrt_distances = None;
|
|
141
|
|
|
_capture_objects = None; # Store indexes of input points that were captured by each neurons individually at the end.
|
|
142
|
|
|
_neighbors = None; # Indexes of neighbours for each neuron.
|
|
143
|
|
|
|
|
144
|
|
|
# describe learning process and internal state
|
|
145
|
|
|
_epochs = 0; # Iteration for learning.
|
|
146
|
|
|
_params = None;
|
|
147
|
|
|
|
|
148
|
|
|
# dynamic changes learning parameters
|
|
149
|
|
|
_local_radius = 0.0;
|
|
150
|
|
|
_learn_rate = 0.0;
|
|
151
|
|
|
|
|
152
|
|
|
__ccore_som_pointer = None;
|
|
153
|
|
|
|
|
154
|
|
|
|
|
155
|
|
|
@property
|
|
156
|
|
|
def size(self):
|
|
157
|
|
|
"""!
|
|
158
|
|
|
@return (uint) Size of self-organized map (number of neurons).
|
|
159
|
|
|
|
|
160
|
|
|
"""
|
|
161
|
|
|
|
|
162
|
|
|
if (self.__ccore_som_pointer is not None):
|
|
163
|
|
|
self._size = wrapper.som_get_size(self.__ccore_som_pointer);
|
|
164
|
|
|
|
|
165
|
|
|
return self._size;
|
|
166
|
|
|
|
|
167
|
|
|
@property
|
|
168
|
|
|
def weights(self):
|
|
169
|
|
|
"""!
|
|
170
|
|
|
@return (list) Weights of each neuron.
|
|
171
|
|
|
|
|
172
|
|
|
"""
|
|
173
|
|
|
|
|
174
|
|
|
if (self.__ccore_som_pointer is not None):
|
|
175
|
|
|
self._weights = wrapper.som_get_weights(self.__ccore_som_pointer);
|
|
176
|
|
|
|
|
177
|
|
|
return self._weights;
|
|
178
|
|
|
|
|
179
|
|
|
@property
|
|
180
|
|
|
def awards(self):
|
|
181
|
|
|
"""!
|
|
182
|
|
|
@return (list) Numbers of captured objects by each neuron.
|
|
183
|
|
|
|
|
184
|
|
|
"""
|
|
185
|
|
|
|
|
186
|
|
|
if (self.__ccore_som_pointer is not None):
|
|
187
|
|
|
self._award = wrapper.som_get_awards(self.__ccore_som_pointer);
|
|
188
|
|
|
|
|
189
|
|
|
return self._award;
|
|
190
|
|
|
|
|
191
|
|
|
@property
|
|
192
|
|
|
def capture_objects(self):
|
|
193
|
|
|
"""!
|
|
194
|
|
|
@return (list) Indexes of captured objects by each neuron.
|
|
195
|
|
|
|
|
196
|
|
|
"""
|
|
197
|
|
|
|
|
198
|
|
|
if (self.__ccore_som_pointer is not None):
|
|
199
|
|
|
self._capture_objects = wrapper.som_get_capture_objects(self.__ccore_som_pointer);
|
|
200
|
|
|
|
|
201
|
|
|
return self._capture_objects;
|
|
202
|
|
|
|
|
203
|
|
|
|
|
204
|
|
|
def __init__(self, rows, cols, conn_type = type_conn.grid_eight, parameters = None, ccore = False):
|
|
205
|
|
|
"""!
|
|
206
|
|
|
@brief Constructor of self-organized map.
|
|
207
|
|
|
|
|
208
|
|
|
@param[in] rows (uint): Number of neurons in the column (number of rows).
|
|
209
|
|
|
@param[in] cols (uint): Number of neurons in the row (number of columns).
|
|
210
|
|
|
@param[in] conn_type (type_conn): Type of connection between oscillators in the network (grid four, grid eight, honeycomb, function neighbour).
|
|
211
|
|
|
@param[in] parameters (som_parameters): Other specific parameters.
|
|
212
|
|
|
@param[in] ccore (bool): If True simulation is performed by CCORE library (C++ implementation of pyclustering).
|
|
213
|
|
|
|
|
214
|
|
|
"""
|
|
215
|
|
|
|
|
216
|
|
|
# some of these parameters are required despite core implementation, for example, for network demonstration.
|
|
217
|
|
|
self._cols = cols;
|
|
218
|
|
|
self._rows = rows;
|
|
219
|
|
|
self._size = cols * rows;
|
|
220
|
|
|
self._conn_type = conn_type;
|
|
221
|
|
|
|
|
222
|
|
|
if (parameters is not None):
|
|
223
|
|
|
self._params = parameters;
|
|
224
|
|
|
else:
|
|
225
|
|
|
self._params = som_parameters();
|
|
226
|
|
|
|
|
227
|
|
|
if (self._params.init_radius is None):
|
|
228
|
|
|
if ((cols + rows) / 4.0 > 1.0):
|
|
229
|
|
|
self._params.init_radius = 2.0;
|
|
230
|
|
|
elif ( (cols > 1) and (rows > 1) ):
|
|
231
|
|
|
self._params.init_radius = 1.5;
|
|
232
|
|
|
else:
|
|
233
|
|
|
self._params.init_radius = 1.0;
|
|
234
|
|
|
|
|
235
|
|
|
if (ccore is True):
|
|
236
|
|
|
self.__ccore_som_pointer = wrapper.som_create(rows, cols, conn_type, self._params);
|
|
237
|
|
|
|
|
238
|
|
|
else:
|
|
239
|
|
|
# location
|
|
240
|
|
|
self._location = list();
|
|
241
|
|
|
for i in range(self._rows):
|
|
242
|
|
|
for j in range(self._cols):
|
|
243
|
|
|
self._location.append([float(i), float(j)]);
|
|
244
|
|
|
|
|
245
|
|
|
# awards
|
|
246
|
|
|
self._award = [0] * self._size;
|
|
247
|
|
|
self._capture_objects = [ [] for i in range(self._size) ];
|
|
248
|
|
|
|
|
249
|
|
|
# distances
|
|
250
|
|
|
self._sqrt_distances = [ [ [] for i in range(self._size) ] for j in range(self._size) ];
|
|
251
|
|
|
for i in range(self._size):
|
|
252
|
|
|
for j in range(i, self._size, 1):
|
|
253
|
|
|
dist = euclidean_distance_sqrt(self._location[i], self._location[j]);
|
|
254
|
|
|
self._sqrt_distances[i][j] = dist;
|
|
255
|
|
|
self._sqrt_distances[j][i] = dist;
|
|
256
|
|
|
|
|
257
|
|
|
# connections
|
|
258
|
|
|
if (conn_type != type_conn.func_neighbor):
|
|
259
|
|
|
self._create_connections(conn_type);
|
|
260
|
|
|
|
|
261
|
|
|
|
|
262
|
|
|
def __del__(self):
|
|
263
|
|
|
"""!
|
|
264
|
|
|
@brief Destructor of the self-organized feature map.
|
|
265
|
|
|
|
|
266
|
|
|
"""
|
|
267
|
|
|
|
|
268
|
|
|
if (self.__ccore_som_pointer is not None):
|
|
269
|
|
|
wrapper.som_destroy(self.__ccore_som_pointer);
|
|
270
|
|
|
|
|
271
|
|
|
|
|
272
|
|
|
def __len__(self):
|
|
273
|
|
|
"""!
|
|
274
|
|
|
@return (uint) Size of self-organized map (number of neurons).
|
|
275
|
|
|
|
|
276
|
|
|
"""
|
|
277
|
|
|
|
|
278
|
|
|
return self.size;
|
|
279
|
|
|
|
|
280
|
|
|
|
|
281
|
|
|
def _create_initial_weights(self, init_type):
|
|
282
|
|
|
"""!
|
|
283
|
|
|
@brief Creates initial weights for neurons in line with the specified initialization.
|
|
284
|
|
|
|
|
285
|
|
|
@param[in] init_type (type_init): Type of initialization of initial neuron weights (random, random in center of the input data, random distributed in data, ditributed in line with uniform grid).
|
|
286
|
|
|
|
|
287
|
|
|
"""
|
|
288
|
|
|
|
|
289
|
|
|
dimension = len(self._data[0]);
|
|
290
|
|
|
|
|
291
|
|
|
maximum_dimension = [self._data[0][i] for i in range(dimension)];
|
|
292
|
|
|
minimum_dimension = [self._data[0][i] for i in range(dimension)];
|
|
293
|
|
|
for i in range(len(self._data)):
|
|
294
|
|
|
for dim in range(dimension):
|
|
295
|
|
|
if (maximum_dimension[dim] < self._data[i][dim]):
|
|
296
|
|
|
maximum_dimension[dim] = self._data[i][dim];
|
|
297
|
|
|
elif (minimum_dimension[dim] > self._data[i][dim]):
|
|
298
|
|
|
minimum_dimension[dim] = self._data[i][dim];
|
|
299
|
|
|
|
|
300
|
|
|
# Increase border
|
|
301
|
|
|
width_dimension = [0] * dimension;
|
|
302
|
|
|
center_dimension = [0] * dimension;
|
|
303
|
|
|
for dim in range(dimension):
|
|
304
|
|
|
width_dimension[dim] = maximum_dimension[dim] - minimum_dimension[dim];
|
|
305
|
|
|
center_dimension[dim] = (maximum_dimension[dim] + minimum_dimension[dim]) / 2;
|
|
306
|
|
|
|
|
307
|
|
|
step_x = center_dimension[0];
|
|
308
|
|
|
if (self._rows > 1): step_x = width_dimension[0] / (self._rows - 1);
|
|
309
|
|
|
|
|
310
|
|
|
step_y = 0.0;
|
|
311
|
|
|
if (dimension > 1):
|
|
312
|
|
|
step_y = center_dimension[1];
|
|
313
|
|
|
if (self._cols > 1): step_y = width_dimension[1] / (self._cols - 1);
|
|
314
|
|
|
|
|
315
|
|
|
# generate weights (topological coordinates)
|
|
316
|
|
|
random.seed();
|
|
317
|
|
|
|
|
318
|
|
|
# Feature SOM 0002: Uniform grid.
|
|
319
|
|
|
if (init_type == type_init.uniform_grid):
|
|
320
|
|
|
# Predefined weights in line with input data.
|
|
321
|
|
|
self._weights = [ [ [] for i in range(dimension) ] for j in range(self._size)];
|
|
322
|
|
|
for i in range(self._size):
|
|
323
|
|
|
location = self._location[i];
|
|
324
|
|
|
for dim in range(dimension):
|
|
325
|
|
|
if (dim == 0):
|
|
326
|
|
|
if (self._rows > 1):
|
|
327
|
|
|
self._weights[i][dim] = minimum_dimension[dim] + step_x * location[dim];
|
|
328
|
|
|
else:
|
|
329
|
|
|
self._weights[i][dim] = center_dimension[dim];
|
|
330
|
|
|
|
|
331
|
|
|
elif (dim == 1):
|
|
332
|
|
|
if (self._cols > 1):
|
|
333
|
|
|
self._weights[i][dim] = minimum_dimension[dim] + step_y * location[dim];
|
|
334
|
|
|
else:
|
|
335
|
|
|
self._weights[i][dim] = center_dimension[dim];
|
|
336
|
|
|
else:
|
|
337
|
|
|
self._weights[i][dim] = center_dimension[dim];
|
|
338
|
|
|
|
|
339
|
|
|
elif (init_type == type_init.random_surface):
|
|
340
|
|
|
# Random weights at the full surface.
|
|
341
|
|
|
self._weights = [ [random.uniform(minimum_dimension[i], maximum_dimension[i]) for i in range(dimension)] for j in range(self._size) ];
|
|
342
|
|
|
|
|
343
|
|
|
elif (init_type == type_init.random_centroid):
|
|
344
|
|
|
# Random weights at the center of input data.
|
|
345
|
|
|
self._weights = [ [(random.random() + center_dimension[i]) for i in range(dimension)] for j in range(self._size) ];
|
|
346
|
|
|
|
|
347
|
|
|
else:
|
|
348
|
|
|
# Random weights of input data.
|
|
349
|
|
|
self._weights = [ [random.random() for i in range(dimension)] for j in range(self._size) ];
|
|
350
|
|
|
|
|
351
|
|
|
def _create_connections(self, conn_type):
|
|
352
|
|
|
"""!
|
|
353
|
|
|
@brief Create connections in line with input rule (grid four, grid eight, honeycomb, function neighbour).
|
|
354
|
|
|
|
|
355
|
|
|
@param[in] conn_type (type_conn): Type of connection between oscillators in the network.
|
|
356
|
|
|
|
|
357
|
|
|
"""
|
|
358
|
|
|
|
|
359
|
|
|
self._neighbors = [[] for index in range(self._size)];
|
|
360
|
|
|
|
|
361
|
|
|
for index in range(0, self._size, 1):
|
|
362
|
|
|
upper_index = index - self._cols;
|
|
363
|
|
|
upper_left_index = index - self._cols - 1;
|
|
364
|
|
|
upper_right_index = index - self._cols + 1;
|
|
365
|
|
|
|
|
366
|
|
|
lower_index = index + self._cols;
|
|
367
|
|
|
lower_left_index = index + self._cols - 1;
|
|
368
|
|
|
lower_right_index = index + self._cols + 1;
|
|
369
|
|
|
|
|
370
|
|
|
left_index = index - 1;
|
|
371
|
|
|
right_index = index + 1;
|
|
372
|
|
|
|
|
373
|
|
|
node_row_index = math.floor(index / self._cols);
|
|
374
|
|
|
upper_row_index = node_row_index - 1;
|
|
375
|
|
|
lower_row_index = node_row_index + 1;
|
|
376
|
|
|
|
|
377
|
|
|
if ( (conn_type == type_conn.grid_eight) or (conn_type == type_conn.grid_four) ):
|
|
378
|
|
|
if (upper_index >= 0):
|
|
379
|
|
|
self._neighbors[index].append(upper_index);
|
|
380
|
|
|
|
|
381
|
|
|
if (lower_index < self._size):
|
|
382
|
|
|
self._neighbors[index].append(lower_index);
|
|
383
|
|
|
|
|
384
|
|
|
if ( (conn_type == type_conn.grid_eight) or (conn_type == type_conn.grid_four) or (conn_type == type_conn.honeycomb) ):
|
|
385
|
|
|
if ( (left_index >= 0) and (math.floor(left_index / self._cols) == node_row_index) ):
|
|
386
|
|
|
self._neighbors[index].append(left_index);
|
|
387
|
|
|
|
|
388
|
|
|
if ( (right_index < self._size) and (math.floor(right_index / self._cols) == node_row_index) ):
|
|
389
|
|
|
self._neighbors[index].append(right_index);
|
|
390
|
|
|
|
|
391
|
|
|
|
|
392
|
|
|
if (conn_type == type_conn.grid_eight):
|
|
393
|
|
|
if ( (upper_left_index >= 0) and (math.floor(upper_left_index / self._cols) == upper_row_index) ):
|
|
394
|
|
|
self._neighbors[index].append(upper_left_index);
|
|
395
|
|
|
|
|
396
|
|
|
if ( (upper_right_index >= 0) and (math.floor(upper_right_index / self._cols) == upper_row_index) ):
|
|
397
|
|
|
self._neighbors[index].append(upper_right_index);
|
|
398
|
|
|
|
|
399
|
|
|
if ( (lower_left_index < self._size) and (math.floor(lower_left_index / self._cols) == lower_row_index) ):
|
|
400
|
|
|
self._neighbors[index].append(lower_left_index);
|
|
401
|
|
|
|
|
402
|
|
|
if ( (lower_right_index < self._size) and (math.floor(lower_right_index / self._cols) == lower_row_index) ):
|
|
403
|
|
|
self._neighbors[index].append(lower_right_index);
|
|
404
|
|
|
|
|
405
|
|
|
|
|
406
|
|
|
if (conn_type == type_conn.honeycomb):
|
|
407
|
|
|
if ( (node_row_index % 2) == 0):
|
|
408
|
|
|
upper_left_index = index - self._cols;
|
|
409
|
|
|
upper_right_index = index - self._cols + 1;
|
|
410
|
|
|
|
|
411
|
|
|
lower_left_index = index + self._cols;
|
|
412
|
|
|
lower_right_index = index + self._cols + 1;
|
|
413
|
|
|
else:
|
|
414
|
|
|
upper_left_index = index - self._cols - 1;
|
|
415
|
|
|
upper_right_index = index - self._cols;
|
|
416
|
|
|
|
|
417
|
|
|
lower_left_index = index + self._cols - 1;
|
|
418
|
|
|
lower_right_index = index + self._cols;
|
|
419
|
|
|
|
|
420
|
|
|
if ( (upper_left_index >= 0) and (math.floor(upper_left_index / self._cols) == upper_row_index) ):
|
|
421
|
|
|
self._neighbors[index].append(upper_left_index);
|
|
422
|
|
|
|
|
423
|
|
|
if ( (upper_right_index >= 0) and (math.floor(upper_right_index / self._cols) == upper_row_index) ):
|
|
424
|
|
|
self._neighbors[index].append(upper_right_index);
|
|
425
|
|
|
|
|
426
|
|
|
if ( (lower_left_index < self._size) and (math.floor(lower_left_index / self._cols) == lower_row_index) ):
|
|
427
|
|
|
self._neighbors[index].append(lower_left_index);
|
|
428
|
|
|
|
|
429
|
|
|
if ( (lower_right_index < self._size) and (math.floor(lower_right_index / self._cols) == lower_row_index) ):
|
|
430
|
|
|
self._neighbors[index].append(lower_right_index);
|
|
431
|
|
|
|
|
432
|
|
|
|
|
433
|
|
|
def _competition(self, x):
|
|
434
|
|
|
"""!
|
|
435
|
|
|
@brief Calculates neuron winner (distance, neuron index).
|
|
436
|
|
|
|
|
437
|
|
|
@param[in] x (list): Input pattern from the input data set, for example it can be coordinates of point.
|
|
438
|
|
|
|
|
439
|
|
|
@return (uint) Returns index of neuron that is winner.
|
|
440
|
|
|
|
|
441
|
|
|
"""
|
|
442
|
|
|
|
|
443
|
|
|
index = 0;
|
|
444
|
|
|
minimum = euclidean_distance_sqrt(self._weights[0], x);
|
|
445
|
|
|
|
|
446
|
|
|
for i in range(1, self._size, 1):
|
|
447
|
|
|
candidate = euclidean_distance_sqrt(self._weights[i], x);
|
|
448
|
|
|
if (candidate < minimum):
|
|
449
|
|
|
index = i;
|
|
450
|
|
|
minimum = candidate;
|
|
451
|
|
|
|
|
452
|
|
|
return index;
|
|
453
|
|
|
|
|
454
|
|
|
|
|
455
|
|
|
def _adaptation(self, index, x):
|
|
456
|
|
|
"""!
|
|
457
|
|
|
@brief Change weight of neurons in line with won neuron.
|
|
458
|
|
|
|
|
459
|
|
|
@param[in] index (uint): Index of neuron-winner.
|
|
460
|
|
|
@param[in] x (list): Input pattern from the input data set.
|
|
461
|
|
|
|
|
462
|
|
|
"""
|
|
463
|
|
|
|
|
464
|
|
|
dimension = len(self._weights[0]);
|
|
465
|
|
|
|
|
466
|
|
|
if (self._conn_type == type_conn.func_neighbor):
|
|
467
|
|
|
for neuron_index in range(self._size):
|
|
468
|
|
|
distance = self._sqrt_distances[index][neuron_index];
|
|
469
|
|
|
|
|
470
|
|
View Code Duplication |
if (distance < self._local_radius):
|
|
|
|
|
|
|
471
|
|
|
influence = math.exp( -( distance / (2.0 * self._local_radius) ) );
|
|
472
|
|
|
|
|
473
|
|
|
for i in range(dimension):
|
|
474
|
|
|
self._weights[neuron_index][i] = self._weights[neuron_index][i] + self._learn_rate * influence * (x[i] - self._weights[neuron_index][i]);
|
|
475
|
|
|
|
|
476
|
|
|
else:
|
|
477
|
|
|
for i in range(dimension):
|
|
478
|
|
|
self._weights[index][i] = self._weights[index][i] + self._learn_rate * (x[i] - self._weights[index][i]);
|
|
479
|
|
|
|
|
480
|
|
|
for neighbor_index in self._neighbors[index]:
|
|
481
|
|
|
distance = self._sqrt_distances[index][neighbor_index]
|
|
482
|
|
View Code Duplication |
if (distance < self._local_radius):
|
|
|
|
|
|
|
483
|
|
|
influence = math.exp( -( distance / (2.0 * self._local_radius) ) );
|
|
484
|
|
|
|
|
485
|
|
|
for i in range(dimension):
|
|
486
|
|
|
self._weights[neighbor_index][i] = self._weights[neighbor_index][i] + self._learn_rate * influence * (x[i] - self._weights[neighbor_index][i]);
|
|
487
|
|
|
|
|
488
|
|
|
|
|
489
|
|
|
def train(self, data, epochs, autostop = False):
|
|
490
|
|
|
"""!
|
|
491
|
|
|
@brief Trains self-organized feature map (SOM).
|
|
492
|
|
|
|
|
493
|
|
|
@param[in] data (list): Input data - list of points where each point is represented by list of features, for example coordinates.
|
|
494
|
|
|
@param[in] epochs (uint): Number of epochs for training.
|
|
495
|
|
|
@param[in] autostop (bool): Automatic termination of learining process when adaptation is not occurred.
|
|
496
|
|
|
|
|
497
|
|
|
@return (uint) Number of learining iterations.
|
|
498
|
|
|
|
|
499
|
|
|
"""
|
|
500
|
|
|
|
|
501
|
|
|
if (self.__ccore_som_pointer is not None):
|
|
502
|
|
|
return wrapper.som_train(self.__ccore_som_pointer, data, epochs, autostop);
|
|
503
|
|
|
|
|
504
|
|
|
for i in range(self._size):
|
|
505
|
|
|
self._award[i] = 0;
|
|
506
|
|
|
self._capture_objects[i].clear();
|
|
507
|
|
|
|
|
508
|
|
|
self._epochs = epochs;
|
|
509
|
|
|
self._data = data;
|
|
510
|
|
|
|
|
511
|
|
|
# weights
|
|
512
|
|
|
self._create_initial_weights(self._params.init_type);
|
|
513
|
|
|
|
|
514
|
|
|
previous_weights = None;
|
|
515
|
|
|
|
|
516
|
|
|
for epoch in range(1, self._epochs + 1):
|
|
517
|
|
|
# Depression term of coupling
|
|
518
|
|
|
self._local_radius = ( self._params.init_radius * math.exp(-(epoch / self._epochs)) ) ** 2;
|
|
519
|
|
|
self._learn_rate = self._params.init_learn_rate * math.exp(-(epoch / self._epochs));
|
|
520
|
|
|
|
|
521
|
|
|
#random.shuffle(self._data); # Random order
|
|
522
|
|
|
|
|
523
|
|
|
# Feature SOM 0003: Clear statistics
|
|
524
|
|
|
if (autostop == True):
|
|
525
|
|
|
for i in range(self._size):
|
|
526
|
|
|
self._award[i] = 0;
|
|
527
|
|
|
self._capture_objects[i].clear();
|
|
528
|
|
|
|
|
529
|
|
|
for i in range(len(self._data)):
|
|
530
|
|
|
# Step 1: Competition:
|
|
531
|
|
|
index = self._competition(self._data[i]);
|
|
532
|
|
|
|
|
533
|
|
|
# Step 2: Adaptation:
|
|
534
|
|
|
self._adaptation(index, self._data[i]);
|
|
535
|
|
|
|
|
536
|
|
|
# Update statistics
|
|
537
|
|
|
if ( (autostop == True) or (epoch == self._epochs) ):
|
|
538
|
|
|
self._award[index] += 1;
|
|
539
|
|
|
self._capture_objects[index].append(i);
|
|
540
|
|
|
|
|
541
|
|
|
# Feature SOM 0003: Check requirement of stopping
|
|
542
|
|
|
if (autostop == True):
|
|
543
|
|
|
if (previous_weights is not None):
|
|
544
|
|
|
maximal_adaptation = self._get_maximal_adaptation(previous_weights);
|
|
545
|
|
|
if (maximal_adaptation < self._params.adaptation_threshold):
|
|
546
|
|
|
return epoch;
|
|
547
|
|
|
|
|
548
|
|
|
previous_weights = [item[:] for item in self._weights];
|
|
549
|
|
|
|
|
550
|
|
|
return self._epochs;
|
|
551
|
|
|
|
|
552
|
|
|
def simulate(self, input_pattern):
|
|
553
|
|
|
"""!
|
|
554
|
|
|
@brief Processes input pattern (no learining) and returns index of neuron-winner.
|
|
555
|
|
|
Using index of neuron winner catched object can be obtained using property capture_objects.
|
|
556
|
|
|
|
|
557
|
|
|
@param[in] input_pattern (list): Input pattern.
|
|
558
|
|
|
|
|
559
|
|
|
@return (uint) Returns index of neuron-winner.
|
|
560
|
|
|
|
|
561
|
|
|
@see capture_objects
|
|
562
|
|
|
|
|
563
|
|
|
"""
|
|
564
|
|
|
|
|
565
|
|
|
if (self.__ccore_som_pointer is not None):
|
|
566
|
|
|
return wrapper.som_simulate(self.__ccore_som_pointer, [ input_pattern ]);
|
|
567
|
|
|
|
|
568
|
|
|
return self._competition(input_pattern);
|
|
569
|
|
|
|
|
570
|
|
|
|
|
571
|
|
|
def _get_maximal_adaptation(self, previous_weights):
|
|
572
|
|
|
"""!
|
|
573
|
|
|
@brief Calculates maximum changes of weight in line with comparison between previous weights and current weights.
|
|
574
|
|
|
|
|
575
|
|
|
@param[in] previous_weights (list): Weights from the previous step of learning process.
|
|
576
|
|
|
|
|
577
|
|
|
@return (double) Value that represents maximum changes of weight after adaptation process.
|
|
578
|
|
|
|
|
579
|
|
|
"""
|
|
580
|
|
|
|
|
581
|
|
|
dimension = len(self._data[0]);
|
|
582
|
|
|
maximal_adaptation = 0.0;
|
|
583
|
|
|
|
|
584
|
|
|
for neuron_index in range(self._size):
|
|
585
|
|
|
for dim in range(dimension):
|
|
586
|
|
|
current_adaptation = previous_weights[neuron_index][dim] - self._weights[neuron_index][dim];
|
|
587
|
|
|
|
|
588
|
|
|
if (current_adaptation < 0): current_adaptation = -current_adaptation;
|
|
589
|
|
|
|
|
590
|
|
|
if (maximal_adaptation < current_adaptation):
|
|
591
|
|
|
maximal_adaptation = current_adaptation;
|
|
592
|
|
|
|
|
593
|
|
|
return maximal_adaptation;
|
|
594
|
|
|
|
|
595
|
|
|
|
|
596
|
|
|
def get_winner_number(self):
|
|
597
|
|
|
"""!
|
|
598
|
|
|
@brief Calculates number of winner at the last step of learning process.
|
|
599
|
|
|
|
|
600
|
|
|
@return (uint) Number of winner.
|
|
601
|
|
|
|
|
602
|
|
|
"""
|
|
603
|
|
|
|
|
604
|
|
|
if (self.__ccore_som_pointer is not None):
|
|
605
|
|
|
self._award = wrapper.som_get_awards(self.__ccore_som_pointer);
|
|
606
|
|
|
|
|
607
|
|
|
winner_number = 0;
|
|
608
|
|
|
for i in range(self._size):
|
|
609
|
|
|
if (self._award[i] > 0):
|
|
610
|
|
|
winner_number += 1;
|
|
611
|
|
|
|
|
612
|
|
|
return winner_number;
|
|
613
|
|
|
|
|
614
|
|
|
|
|
615
|
|
|
def show_distance_matrix(self):
|
|
616
|
|
|
"""!
|
|
617
|
|
|
@brief Shows gray visualization of U-matrix (distance matrix).
|
|
618
|
|
|
|
|
619
|
|
|
@see get_distance_matrix()
|
|
620
|
|
|
|
|
621
|
|
|
"""
|
|
622
|
|
|
distance_matrix = self.get_distance_matrix();
|
|
623
|
|
|
|
|
624
|
|
|
plt.imshow(distance_matrix, cmap = plt.get_cmap('hot'), interpolation='kaiser');
|
|
625
|
|
|
plt.title("U-Matrix");
|
|
626
|
|
|
plt.colorbar();
|
|
627
|
|
|
plt.show();
|
|
628
|
|
|
|
|
629
|
|
|
|
|
630
|
|
|
def get_distance_matrix(self):
|
|
631
|
|
|
"""!
|
|
632
|
|
|
@brief Calculates distance matrix (U-matrix).
|
|
633
|
|
|
@details The U-Matrix visualizes based on the distance in input space between a weight vector and its neighbors on map.
|
|
634
|
|
|
|
|
635
|
|
|
@return (list) Distance matrix (U-matrix).
|
|
636
|
|
|
|
|
637
|
|
|
@see show_distance_matrix()
|
|
638
|
|
|
@see get_density_matrix()
|
|
639
|
|
|
|
|
640
|
|
|
"""
|
|
641
|
|
|
if (self.__ccore_som_pointer is not None):
|
|
642
|
|
|
self._weights = wrapper.som_get_weights(self.__ccore_som_pointer);
|
|
643
|
|
|
|
|
644
|
|
|
if (self._conn_type != type_conn.func_neighbor):
|
|
645
|
|
|
self._neighbors = wrapper.som_get_neighbors(self.__ccore_som_pointer);
|
|
646
|
|
|
|
|
647
|
|
|
distance_matrix = [ [0.0] * self._cols for i in range(self._rows) ];
|
|
648
|
|
|
|
|
649
|
|
|
for i in range(self._rows):
|
|
650
|
|
|
for j in range(self._cols):
|
|
651
|
|
|
neuron_index = i * self._cols + j;
|
|
652
|
|
|
|
|
653
|
|
|
if (self._conn_type == type_conn.func_neighbor):
|
|
654
|
|
|
self._create_connections(type_conn.grid_eight);
|
|
655
|
|
|
|
|
656
|
|
|
for neighbor_index in self._neighbors[neuron_index]:
|
|
657
|
|
|
distance_matrix[i][j] += euclidean_distance_sqrt(self._weights[neuron_index], self._weights[neighbor_index]);
|
|
658
|
|
|
|
|
659
|
|
|
distance_matrix[i][j] /= len(self._neighbors[neuron_index]);
|
|
660
|
|
|
|
|
661
|
|
|
return distance_matrix;
|
|
662
|
|
|
|
|
663
|
|
|
|
|
664
|
|
|
def show_density_matrix(self, surface_divider = 20.0):
|
|
|
|
|
|
|
665
|
|
|
"""!
|
|
666
|
|
|
@brief Show density matrix (P-matrix) using kernel density estimation.
|
|
667
|
|
|
|
|
668
|
|
|
@param[in] surface_divider (double): Divider in each dimension that affect radius for density measurement.
|
|
669
|
|
|
|
|
670
|
|
|
@see show_distance_matrix()
|
|
671
|
|
|
|
|
672
|
|
|
"""
|
|
673
|
|
|
density_matrix = self.get_density_matrix();
|
|
674
|
|
|
|
|
675
|
|
|
plt.imshow(density_matrix, cmap = plt.get_cmap('hot'), interpolation='kaiser');
|
|
676
|
|
|
plt.title("P-Matrix");
|
|
677
|
|
|
plt.colorbar();
|
|
678
|
|
|
plt.show();
|
|
679
|
|
|
|
|
680
|
|
|
|
|
681
|
|
|
def get_density_matrix(self, surface_divider = 20.0):
|
|
682
|
|
|
"""!
|
|
683
|
|
|
@brief Calculates density matrix (P-Matrix).
|
|
684
|
|
|
|
|
685
|
|
|
@param[in] surface_divider (double): Divider in each dimension that affect radius for density measurement.
|
|
686
|
|
|
|
|
687
|
|
|
@return (list) Density matrix (P-Matrix).
|
|
688
|
|
|
|
|
689
|
|
|
@see get_distance_matrix()
|
|
690
|
|
|
|
|
691
|
|
|
"""
|
|
692
|
|
|
|
|
693
|
|
|
if (self.__ccore_som_pointer is not None):
|
|
694
|
|
|
self._weights = wrapper.som_get_weights(self.__ccore_som_pointer);
|
|
695
|
|
|
|
|
696
|
|
|
density_matrix = [ [0] * self._cols for i in range(self._rows) ];
|
|
697
|
|
|
dimension = len(self._weights[0]);
|
|
698
|
|
|
|
|
699
|
|
|
dim_max = [ float('-Inf') ] * dimension;
|
|
700
|
|
|
dim_min = [ float('Inf') ] * dimension;
|
|
701
|
|
|
|
|
702
|
|
|
for weight in self._weights:
|
|
703
|
|
|
for index_dim in range(dimension):
|
|
704
|
|
|
if (weight[index_dim] > dim_max[index_dim]):
|
|
705
|
|
|
dim_max[index_dim] = weight[index_dim];
|
|
706
|
|
|
|
|
707
|
|
|
if (weight[index_dim] < dim_min[index_dim]):
|
|
708
|
|
|
dim_min[index_dim] = weight[index_dim];
|
|
709
|
|
|
|
|
710
|
|
|
radius = [0.0] * len(self._weights[0]);
|
|
711
|
|
|
for index_dim in range(dimension):
|
|
712
|
|
|
radius[index_dim] = ( dim_max[index_dim] - dim_min[index_dim] ) / surface_divider;
|
|
713
|
|
|
|
|
714
|
|
|
for point in self._data:
|
|
715
|
|
|
for index_neuron in range(len(self)):
|
|
716
|
|
|
point_covered = True;
|
|
717
|
|
|
|
|
718
|
|
|
for index_dim in range(dimension):
|
|
719
|
|
|
if (abs(point[index_dim] - self._weights[index_neuron][index_dim]) > radius[index_dim]):
|
|
720
|
|
|
point_covered = False;
|
|
721
|
|
|
break;
|
|
722
|
|
|
|
|
723
|
|
|
row = math.floor(index_neuron / self._cols);
|
|
724
|
|
|
col = index_neuron - row * self._cols;
|
|
725
|
|
|
|
|
726
|
|
|
if (point_covered is True):
|
|
727
|
|
|
density_matrix[row][col] += 1;
|
|
728
|
|
|
|
|
729
|
|
|
return density_matrix;
|
|
730
|
|
|
|
|
731
|
|
|
|
|
732
|
|
|
def show_winner_matrix(self):
|
|
733
|
|
|
"""!
|
|
734
|
|
|
@brief Show winner matrix where each element corresponds to neuron and value represents
|
|
735
|
|
|
amount of won objects from input dataspace at the last training iteration.
|
|
736
|
|
|
|
|
737
|
|
|
@see show_distance_matrix()
|
|
738
|
|
|
|
|
739
|
|
|
"""
|
|
740
|
|
|
|
|
741
|
|
|
if (self.__ccore_som_pointer is not None):
|
|
742
|
|
|
self._award = wrapper.som_get_awards(self.__ccore_som_pointer);
|
|
743
|
|
|
|
|
744
|
|
|
(fig, ax) = plt.subplots();
|
|
|
|
|
|
|
745
|
|
|
winner_matrix = [ [0] * self._cols for i in range(self._rows) ];
|
|
746
|
|
|
|
|
747
|
|
|
for i in range(self._rows):
|
|
748
|
|
|
for j in range(self._cols):
|
|
749
|
|
|
neuron_index = i * self._cols + j;
|
|
750
|
|
|
|
|
751
|
|
|
winner_matrix[i][j] = self._award[neuron_index];
|
|
752
|
|
|
ax.text(i, j, str(winner_matrix[i][j]), va='center', ha='center')
|
|
753
|
|
|
|
|
754
|
|
|
ax.imshow(winner_matrix, cmap = plt.get_cmap('cool'), interpolation='none');
|
|
755
|
|
|
ax.grid(True);
|
|
756
|
|
|
|
|
757
|
|
|
plt.title("Winner Matrix");
|
|
758
|
|
|
plt.show();
|
|
759
|
|
|
|
|
760
|
|
|
|
|
761
|
|
|
def show_network(self, awards = False, belongs = False, coupling = True, dataset = True, marker_type = 'o'):
|
|
762
|
|
|
"""!
|
|
763
|
|
|
@brief Shows neurons in the dimension of data.
|
|
764
|
|
|
|
|
765
|
|
|
@param[in] awards (bool): If True - displays how many objects won each neuron.
|
|
766
|
|
|
@param[in] belongs (bool): If True - marks each won object by according index of neuron-winner (only when dataset is displayed too).
|
|
767
|
|
|
@param[in] coupling (bool): If True - displays connections between neurons (except case when function neighbor is used).
|
|
768
|
|
|
@param[in] dataset (bool): If True - displays inputs data set.
|
|
769
|
|
|
@param[in] marker_type (string): Defines marker that is used for dispaying neurons in the network.
|
|
770
|
|
|
|
|
771
|
|
|
"""
|
|
772
|
|
|
|
|
773
|
|
|
if (self.__ccore_som_pointer is not None):
|
|
774
|
|
|
self._size = wrapper.som_get_size(self.__ccore_som_pointer);
|
|
775
|
|
|
self._weights = wrapper.som_get_weights(self.__ccore_som_pointer);
|
|
776
|
|
|
self._neighbors = wrapper.som_get_neighbors(self.__ccore_som_pointer);
|
|
777
|
|
|
self._award = wrapper.som_get_awards(self.__ccore_som_pointer);
|
|
778
|
|
|
|
|
779
|
|
|
dimension = len(self._weights[0]);
|
|
780
|
|
|
|
|
781
|
|
|
fig = plt.figure();
|
|
782
|
|
|
axes = None;
|
|
783
|
|
|
|
|
784
|
|
|
# Check for dimensions
|
|
785
|
|
|
if ( (dimension == 1) or (dimension == 2) ):
|
|
786
|
|
|
axes = fig.add_subplot(111);
|
|
787
|
|
|
elif (dimension == 3):
|
|
788
|
|
|
axes = fig.gca(projection='3d');
|
|
789
|
|
|
else:
|
|
790
|
|
|
raise NameError('Dwawer supports only 1D, 2D and 3D data representation');
|
|
791
|
|
|
|
|
792
|
|
|
|
|
793
|
|
|
# Show data
|
|
794
|
|
|
if (dataset == True):
|
|
795
|
|
|
for x in self._data:
|
|
796
|
|
|
if (dimension == 1):
|
|
797
|
|
|
axes.plot(x[0], 0.0, 'b|', ms = 30);
|
|
798
|
|
|
|
|
799
|
|
|
elif (dimension == 2):
|
|
800
|
|
|
axes.plot(x[0], x[1], 'b.');
|
|
801
|
|
|
|
|
802
|
|
|
elif (dimension == 3):
|
|
803
|
|
|
axes.scatter(x[0], x[1], x[2], c = 'b', marker = '.');
|
|
804
|
|
|
|
|
805
|
|
|
# Show neurons
|
|
806
|
|
|
for index in range(self._size):
|
|
807
|
|
|
color = 'g';
|
|
808
|
|
|
if (self._award[index] == 0): color = 'y';
|
|
809
|
|
|
|
|
810
|
|
|
if (dimension == 1):
|
|
811
|
|
|
axes.plot(self._weights[index][0], 0.0, color + marker_type);
|
|
812
|
|
|
|
|
813
|
|
|
if (awards == True):
|
|
814
|
|
|
location = '{0}'.format(self._award[index]);
|
|
815
|
|
|
axes.text(self._weights[index][0], 0.0, location, color='black', fontsize = 10);
|
|
816
|
|
|
|
|
817
|
|
|
if (belongs == True):
|
|
818
|
|
|
location = '{0}'.format(index);
|
|
819
|
|
|
axes.text(self._weights[index][0], 0.0, location, color='black', fontsize = 12);
|
|
820
|
|
|
for k in range(len(self._capture_objects[index])):
|
|
821
|
|
|
point = self._data[self._capture_objects[index][k]];
|
|
822
|
|
|
axes.text(point[0], 0.0, location, color='blue', fontsize = 10);
|
|
823
|
|
|
|
|
824
|
|
|
if (dimension == 2):
|
|
825
|
|
|
axes.plot(self._weights[index][0], self._weights[index][1], color + marker_type);
|
|
826
|
|
|
|
|
827
|
|
|
if (awards == True):
|
|
828
|
|
|
location = '{0}'.format(self._award[index]);
|
|
829
|
|
|
axes.text(self._weights[index][0], self._weights[index][1], location, color='black', fontsize = 10);
|
|
830
|
|
|
|
|
831
|
|
|
if (belongs == True):
|
|
832
|
|
|
location = '{0}'.format(index);
|
|
833
|
|
|
axes.text(self._weights[index][0], self._weights[index][1], location, color='black', fontsize = 12);
|
|
834
|
|
|
for k in range(len(self._capture_objects[index])):
|
|
835
|
|
|
point = self._data[self._capture_objects[index][k]];
|
|
836
|
|
|
axes.text(point[0], point[1], location, color='blue', fontsize = 10);
|
|
837
|
|
|
|
|
838
|
|
|
if ( (self._conn_type != type_conn.func_neighbor) and (coupling != False) ):
|
|
839
|
|
|
for neighbor in self._neighbors[index]:
|
|
840
|
|
|
if (neighbor > index):
|
|
841
|
|
|
axes.plot([self._weights[index][0], self._weights[neighbor][0]], [self._weights[index][1], self._weights[neighbor][1]], 'g', linewidth = 0.5);
|
|
842
|
|
|
|
|
843
|
|
|
elif (dimension == 3):
|
|
844
|
|
|
axes.scatter(self._weights[index][0], self._weights[index][1], self._weights[index][2], c = color, marker = marker_type);
|
|
845
|
|
|
|
|
846
|
|
|
if ( (self._conn_type != type_conn.func_neighbor) and (coupling != False) ):
|
|
847
|
|
|
for neighbor in self._neighbors[index]:
|
|
848
|
|
|
if (neighbor > index):
|
|
849
|
|
|
axes.plot([self._weights[index][0], self._weights[neighbor][0]], [self._weights[index][1], self._weights[neighbor][1]], [self._weights[index][2], self._weights[neighbor][2]], 'g-', linewidth = 0.5);
|
|
850
|
|
|
|
|
851
|
|
|
|
|
852
|
|
|
plt.title("Network Structure");
|
|
853
|
|
|
plt.grid();
|
|
854
|
|
|
plt.show(); |
annoviko /
pyclustering
Loading history...
This can be caused by one of the following:
1. Missing Dependencies
This error could indicate a configuration issue of Pylint. Make sure that your libraries are available by adding the necessary commands.
2. Missing __init__.py files
This error could also result from missing
__init__.pyfiles in your module folders. Make sure that you place one file in each sub-folder.