1
|
1 # -*- coding: utf-8 -*-
|
|
2 import sys
|
|
3 if sys.version_info <= (2, 8):
|
|
4 from builtins import super
|
|
5
|
|
6 import os
|
|
7 import codecs
|
|
8 from collections import defaultdict
|
|
9
|
|
10 import numpy as np
|
|
11 import scipy as sp
|
|
12 import pandas as pd
|
|
13 import networkx as nx
|
|
14
|
|
15
|
|
16 __all__ = ["FrozenClass",
|
|
17 "Singleton",
|
|
18 "to_networkx_digraph",
|
|
19 "normalize",
|
|
20 "rand_swap",
|
|
21 "rand_flip",
|
|
22 "rand_weights",
|
|
23 "rand_structure",
|
|
24 "get_akey",
|
|
25 "get_avalue",]
|
|
26
|
|
27
|
|
28 class FrozenClass(object):
|
|
29
|
|
30 __isfrozen = False
|
|
31
|
|
32 def __setattr__(self, key, value):
|
|
33 if self.__isfrozen and not hasattr(self, key):
|
|
34 raise TypeError( "%r is a frozen class" % self )
|
|
35 object.__setattr__(self, key, value)
|
|
36
|
|
37 def _freeze(self):
|
|
38 self.__isfrozen = True
|
|
39
|
|
40 def _melt(self):
|
|
41 self.__isfrozen = False
|
|
42
|
|
43 """
|
|
44 <Reference>
|
|
45 http://stackoverflow.com/questions/3603502/prevent-creating-new-attributes-outside-init
|
|
46 """
|
|
47 # end of def FrozenClass
|
|
48
|
|
49
|
|
50 def Singleton(_class):
|
|
51 class __Singleton(_class):
|
|
52 __instance = None
|
|
53
|
|
54 def __new__(cls, *args, **kwargs):
|
|
55 if cls.__instance is None:
|
|
56 cls.__instance = super().__new__(cls, *args, **kwargs)
|
|
57
|
|
58 # Creation and initialization of '__initialized'
|
|
59 cls.__instance.__initialized = False
|
|
60 # end of if
|
|
61 return cls.__instance
|
|
62
|
|
63 def __init__(self, *args, **kwargs):
|
|
64 if self.__initialized:
|
|
65 return
|
|
66
|
|
67 super().__init__(*args, **kwargs)
|
|
68 self.__initialized = True
|
|
69
|
|
70 def __repr__(self):
|
|
71 return '<{0} Singleton object at {1}>'.format(
|
|
72 _class.__name__, hex(id(self)))
|
|
73
|
|
74 def __str__(self):
|
|
75 return super().__str__()
|
|
76 # end of def class
|
|
77
|
|
78 __Singleton.__name__ = _class.__name__
|
|
79 return __Singleton
|
|
80
|
|
81 """
|
|
82 <References>
|
|
83 http://m.egloos.zum.com/mataeoh/v/7081556
|
|
84 """
|
|
85 # end of def Singleton
|
|
86
|
|
87 def normalize(A, norm_in=True, norm_out=True):
|
|
88 # Check whether A is a square matrix
|
|
89 if A.shape[0] != A.shape[1]:
|
|
90 raise ValueError(
|
|
91 "The A (adjacency matrix) should be square matrix.")
|
|
92
|
|
93 # Build propagation matrix (aka. transition matrix) _W from A
|
|
94 W = A.copy()
|
|
95
|
|
96 # Norm. out-degree
|
|
97 if norm_out == True:
|
|
98 sum_col_A = np.abs(A).sum(axis=0)
|
|
99 sum_col_A[sum_col_A == 0] = 1
|
|
100 if norm_in == False:
|
|
101 Dc = 1 / sum_col_A
|
|
102 else:
|
|
103 Dc = 1 / np.sqrt(sum_col_A)
|
|
104 # end of else
|
|
105 W = Dc * W # This is not matrix multiplication
|
|
106
|
|
107 # Norm. in-degree
|
|
108 if norm_in == True:
|
|
109 sum_row_A = np.abs(A).sum(axis=1)
|
|
110 sum_row_A[sum_row_A == 0] = 1
|
|
111 if norm_out == False:
|
|
112 Dr = 1 / sum_row_A
|
|
113 else:
|
|
114 Dr = 1 / np.sqrt(sum_row_A)
|
|
115 # end of row
|
|
116 W = np.multiply(W, np.mat(Dr).T)
|
|
117 # Converting np.mat to ndarray
|
|
118 # does not cost a lot.
|
|
119 W = W.A
|
|
120 # end of if
|
|
121 """
|
|
122 The normalization above is the same as the follows:
|
|
123 >>> np.diag(Dr).dot(A.dot(np.diag(Dc)))
|
|
124 """
|
|
125 return W
|
|
126
|
|
127
|
|
128 # end of def normalize
|
|
129
|
|
130 def to_networkx_digraph(A, n2i=None):
|
|
131 if not n2i:
|
|
132 return nx.from_numpy_array(A, create_using=nx.Digraph)
|
|
133
|
|
134 i2n = {ix:name for name, ix in n2i.items()}
|
|
135 dg = nx.DiGraph()
|
|
136 ind_row, ind_col = A.nonzero()
|
|
137 for ix_trg, ix_src in zip(ind_row, ind_col):
|
|
138 name_src = i2n[ix_src]
|
|
139 name_trg = i2n[ix_trg]
|
|
140 sign = np.sign(A[ix_trg, ix_src])
|
|
141 dg.add_edge(name_src, name_trg)
|
|
142 dg.edges[name_src, name_trg]['SIGN'] = sign
|
|
143 # end of for
|
|
144 return dg
|
|
145 # end of for
|
|
146 # end of def to_networkx_digraph
|
|
147
|
|
148 def rand_swap(A, nsamp=10, noself=True, pivots=None, inplace=False):
|
|
149 """Randomly rewire the network connections by swapping.
|
|
150
|
|
151 Parameters
|
|
152 ----------
|
|
153 A : numpy.ndarray
|
|
154 Adjacency matrix (connection matrix).
|
|
155 nsamp : int, optional
|
|
156 Number of sampled connections to rewire
|
|
157 noself : bool, optional
|
|
158 Whether to allow self-loop link.
|
|
159 pivots : list, optional
|
|
160 Indices of pivot nodes
|
|
161 inplace : bool, optional
|
|
162 Modify the given adjacency matrix for rewiring.
|
|
163
|
|
164
|
|
165 Returns
|
|
166 -------
|
|
167 B : numpy.ndarray
|
|
168 The randomized matrix.
|
|
169 The reference of the given W is returned, when inplace is True.
|
|
170 """
|
|
171
|
|
172
|
|
173 if not inplace:
|
|
174 A_org = A
|
|
175 B = A.copy() #np.array(A, dtype=np.float64)
|
|
176 else:
|
|
177 A_org = A.copy() #np.array(A, dtype=np.float64)
|
|
178 B = A
|
|
179
|
|
180 cnt = 0
|
|
181 while cnt < nsamp:
|
|
182 ir, ic = B.nonzero()
|
|
183 if pivots:
|
|
184 if np.random.uniform() < 0.5:
|
|
185 isrc1 = np.random.choice(pivots)
|
|
186 nz = B[:, isrc1].nonzero()[0]
|
|
187 if len(nz) == 0:
|
|
188 continue
|
|
189 itrg1 = np.random.choice(nz)
|
|
190 else:
|
|
191 itrg1 = np.random.choice(pivots)
|
|
192 nz = B[itrg1, :].nonzero()[0]
|
|
193 if len(nz) == 0:
|
|
194 continue
|
|
195 isrc1 = np.random.choice(nz)
|
|
196 # if-else
|
|
197
|
|
198 itrg2, isrc2 = itrg1, isrc1
|
|
199 while isrc1 == isrc2 and itrg1 == itrg2:
|
|
200 i2 = np.random.randint(0, ir.size)
|
|
201 itrg2, isrc2 = ir[i2], ic[i2]
|
|
202 else:
|
|
203 i1, i2 = 0, 0
|
|
204 while i1 == i2:
|
|
205 i1, i2 = np.random.randint(0, ir.size, 2)
|
|
206
|
|
207 itrg1, isrc1 = ir[i1], ic[i1]
|
|
208 itrg2, isrc2 = ir[i2], ic[i2]
|
|
209
|
|
210 if noself:
|
|
211 if itrg2 == isrc1 or itrg1 == isrc2:
|
|
212 continue
|
|
213
|
|
214 # Are the swapped links new?
|
|
215 if B[itrg2, isrc1] == 0 and B[itrg1, isrc2] == 0:
|
|
216 a, b = B[itrg1, isrc1], B[itrg2, isrc2]
|
|
217
|
|
218 # Are the swapped links in the original network?
|
|
219 if A_org[itrg2, isrc1] == a and A_org[itrg1, isrc2] == b:
|
|
220 continue
|
|
221
|
|
222 B[itrg2, isrc1], B[itrg1, isrc2] = a, b
|
|
223 B[itrg1, isrc1], B[itrg2, isrc2] = 0, 0
|
|
224 cnt += 1
|
|
225 else:
|
|
226 continue
|
|
227 # end of while
|
|
228
|
|
229 if not inplace:
|
|
230 return B
|
|
231
|
|
232
|
|
233 def rand_flip(A, nsamp=10, pivots=None, inplace=False):
|
|
234 """Randomly flip the signs of connections.
|
|
235
|
|
236 Parameters
|
|
237 ----------
|
|
238 A : numpy.ndarray
|
|
239 Adjacency matrix (connection matrix).
|
|
240 nsamp : int, optional
|
|
241 Number of sampled connections to be flipped.
|
|
242 pivots : list, optional
|
|
243 Indices of pivot nodes
|
|
244 inplace : bool, optional
|
|
245 Modify the given adjacency matrix for rewiring.
|
|
246
|
|
247 Returns
|
|
248 -------
|
|
249 B : numpy.ndarray
|
|
250 The randomized matrix.
|
|
251 The reference of the given W is returned, when inplace is True.
|
|
252 """
|
|
253 if not inplace:
|
|
254 B = A.copy() #np.array(A, dtype=np.float64)
|
|
255 else:
|
|
256 B = A
|
|
257
|
|
258 ir, ic = B.nonzero()
|
|
259 if pivots:
|
|
260 iflip = np.random.choice(pivots, nsamp)
|
|
261 else:
|
|
262 iflip = np.random.randint(0, ir.size, nsamp)
|
|
263
|
|
264 B[ir[iflip], ic[iflip]] *= -1
|
|
265 return B
|
|
266
|
|
267
|
|
268 def rand_weights(W, lb=-3, ub=3, inplace=False):
|
|
269 """ Randomly sample the weights of connections in W from 10^(lb, ub).
|
|
270
|
|
271 Parameters
|
|
272 ----------
|
|
273 W : numpy.ndarray
|
|
274 Adjacency (connection) or weight matrix.
|
|
275 lb : float, optional
|
|
276 The 10's exponent for lower bound
|
|
277 inplace : bool, optional
|
|
278 Modify the given adjacency matrix for rewiring.
|
|
279
|
|
280 Returns
|
|
281 -------
|
|
282 B : numpy.ndarray
|
|
283 The randomly sampled weight matrix.
|
|
284 The reference of the given W is returned, when inplace is True.
|
|
285 """
|
|
286 if not inplace:
|
|
287 B = np.array(W, dtype=np.float64)
|
|
288 else:
|
|
289 if not np.issubdtype(W.dtype, np.floating):
|
|
290 raise ValueError("W.dtype given to rand_weights should be "
|
|
291 "a float type, not %s"%(W.dtype))
|
|
292
|
|
293 B = W
|
|
294 # end of if-else
|
|
295
|
|
296 ir, ic = B.nonzero()
|
|
297 weights_rand = 10 ** np.random.uniform(lb, ub,
|
|
298 size=(ir.size,))
|
|
299
|
|
300 B[ir, ic] = weights_rand*np.sign(B[ir, ic], dtype=np.float)
|
|
301 """The above code is equal to the following:
|
|
302
|
|
303 for i in range(ir.size):
|
|
304 p, q = ir[i], ic[i]
|
|
305 B[p, q] = weights_rand[i] * np.sign(B[p, q], dtype=np.float)
|
|
306 """
|
|
307 return B
|
|
308
|
|
309
|
|
310 def rand_structure(A, nswap=10, nflip=10, noself=True, pivots=None, inplace=False):
|
|
311 if not inplace:
|
|
312 B = A.copy()
|
|
313 else:
|
|
314 B = A
|
|
315 if nflip > 0:
|
|
316 B = rand_flip(B, nflip, pivots, inplace)
|
|
317 if nswap > 0:
|
|
318 B = rand_swap(B, nswap, noself, pivots, inplace)
|
|
319 return B
|
|
320
|
|
321
|
|
322 def get_akey(d):
|
|
323 """Get a key from a given dictionary.
|
|
324 It returns the first key in d.keys().
|
|
325
|
|
326 Parameters
|
|
327 ----------
|
|
328 d : dict
|
|
329 Dictionary of objects.
|
|
330
|
|
331 Returns
|
|
332 -------
|
|
333 obj : object
|
|
334 First item of iter(d.keys()).
|
|
335 """
|
|
336 return next(iter(d.keys()))
|
|
337
|
|
338
|
|
339 def get_avalue(d):
|
|
340 """Get a value from a given dictionary.
|
|
341 It returns the value designated by sfa.get_akey().
|
|
342
|
|
343 Parameters
|
|
344 ----------
|
|
345 d : dict
|
|
346 Dictionary of objects.
|
|
347
|
|
348 Returns
|
|
349 -------
|
|
350 obj : object
|
|
351 First item of d[iter(d.keys())].
|
|
352 """
|
|
353 akey = next(iter(d.keys()))
|
|
354 return d[akey] |