|
1
|
|
|
# encoding=utf8 |
|
2
|
|
|
# pylint: disable=mixed-indentation, multiple-statements, line-too-long, unused-argument, no-self-use, no-self-use, attribute-defined-outside-init, logging-not-lazy, len-as-condition, singleton-comparison, arguments-differ, bad-continuation, dangerous-default-value, keyword-arg-before-vararg |
|
3
|
|
|
import logging |
|
4
|
|
|
|
|
5
|
|
|
from numpy import random as rand, argmin, argmax, mean, cos, asarray, append, sin |
|
6
|
|
|
from scipy.spatial.distance import euclidean |
|
7
|
|
|
|
|
8
|
|
|
from NiaPy.algorithms.algorithm import Algorithm, Individual, defaultIndividualInit |
|
9
|
|
|
from NiaPy.util.utility import objects2array |
|
10
|
|
|
|
|
11
|
|
|
__all__ = ['DifferentialEvolution', 'DynNpDifferentialEvolution', 'AgingNpDifferentialEvolution', 'CrowdingDifferentialEvolution', 'MultiStrategyDifferentialEvolution', 'DynNpMultiStrategyDifferentialEvolution', 'AgingNpMultiMutationDifferentialEvolution', 'AgingIndividual', 'CrossRand1', 'CrossBest2', 'CrossBest1', 'CrossBest2', 'CrossCurr2Rand1', 'CrossCurr2Best1', 'multiMutations'] |
|
12
|
|
|
|
|
13
|
|
|
logging.basicConfig() |
|
14
|
|
|
logger = logging.getLogger('NiaPy.algorithms.basic') |
|
15
|
|
|
logger.setLevel('INFO') |
|
16
|
|
|
|
|
17
|
|
View Code Duplication |
def CrossRand1(pop, ic, x_b, f, cr, rnd=rand, *args): |
|
|
|
|
|
|
18
|
|
|
r"""Mutation strategy with crossover. |
|
19
|
|
|
|
|
20
|
|
|
Mutation strategy uses three different random individuals from population to perform mutation. |
|
21
|
|
|
|
|
22
|
|
|
Mutation: |
|
23
|
|
|
Name: DE/rand/1 |
|
24
|
|
|
|
|
25
|
|
|
:math:`\mathbf{x}_{r_1, G} + F \cdot (\mathbf{x}_{r_2, G} - \mathbf{x}_{r_3, G}` |
|
26
|
|
|
where :math:`r_1, r_2, r_3` are random indexes representing current population individuals. |
|
27
|
|
|
|
|
28
|
|
|
Crossover: |
|
29
|
|
|
Name: Binomial crossover |
|
30
|
|
|
|
|
31
|
|
|
:math:`\mathbf{x}_{i, G+1} = \begin{cases} \mathbf{u}_{i, G+1}, & \text{if $f(\mathbf{u}_{i, G+1}) \leq f(\mathbf{x}_{i, G})$}, \\ \mathbf{x}_{i, G}, & \text{otherwise}. \end{cases}` |
|
32
|
|
|
|
|
33
|
|
|
Args: |
|
34
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
35
|
|
|
ic (int): Index of individual being mutated. |
|
36
|
|
|
x_b (Individual): Current global best individual. |
|
37
|
|
|
f (float): Scale factor. |
|
38
|
|
|
cr (float): Crossover probability. |
|
39
|
|
|
rnd (mtrand.RandomState): Random generator. |
|
40
|
|
|
*args (list): Additional arguments. |
|
41
|
|
|
|
|
42
|
|
|
Returns: |
|
43
|
|
|
numpy.ndarray: Mutated and mixed individual. |
|
44
|
|
|
""" |
|
45
|
|
|
j = rnd.randint(len(pop[ic])) |
|
46
|
|
|
p = [1 / (len(pop) - 1.0) if i != ic else 0 for i in range(len(pop))] if len(pop) > 3 else None |
|
47
|
|
|
r = rnd.choice(len(pop), 3, replace=not len(pop) >= 3, p=p) |
|
48
|
|
|
x = [pop[r[0]][i] + f * (pop[r[1]][i] - pop[r[2]][i]) if rnd.rand() < cr or i == j else pop[ic][i] for i in range(len(pop[ic]))] |
|
49
|
|
|
return asarray(x) |
|
50
|
|
|
|
|
51
|
|
View Code Duplication |
def CrossBest1(pop, ic, x_b, f, cr, rnd=rand, *args): |
|
|
|
|
|
|
52
|
|
|
r"""Mutation strategy with crossover. |
|
53
|
|
|
|
|
54
|
|
|
Mutation strategy uses two different random individuals from population and global best individual. |
|
55
|
|
|
|
|
56
|
|
|
Mutation: |
|
57
|
|
|
Name: de/best/1 |
|
58
|
|
|
|
|
59
|
|
|
:math:`\mathbf{v}_{i, G} = \mathbf{x}_{best, G} + F \cdot (\mathbf{x}_{r_1, G} - \mathbf{x}_{r_2, G})` |
|
60
|
|
|
where :math:`r_1, r_2` are random indexes representing current population individuals. |
|
61
|
|
|
|
|
62
|
|
|
Crossover: |
|
63
|
|
|
Name: Binomial crossover |
|
64
|
|
|
|
|
65
|
|
|
:math:`\mathbf{x}_{i, G+1} = \begin{cases} \mathbf{u}_{i, G+1}, & \text{if $f(\mathbf{u}_{i, G+1}) \leq f(\mathbf{x}_{i, G})$}, \\ \mathbf{x}_{i, G}, & \text{otherwise}. \end{cases}` |
|
66
|
|
|
|
|
67
|
|
|
args: |
|
68
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
69
|
|
|
ic (int): Index of individual being mutated. |
|
70
|
|
|
x_b (Individual): Current global best individual. |
|
71
|
|
|
f (float): Scale factor. |
|
72
|
|
|
cr (float): Crossover probability. |
|
73
|
|
|
rnd (mtrand.RandomState): Random generator. |
|
74
|
|
|
*args (list): Additional arguments. |
|
75
|
|
|
|
|
76
|
|
|
returns: |
|
77
|
|
|
numpy.ndarray: Mutated and mixed individual. |
|
78
|
|
|
""" |
|
79
|
|
|
j = rnd.randint(len(pop[ic])) |
|
80
|
|
|
p = [1 / (len(pop) - 1.0) if i != ic else 0 for i in range(len(pop))] if len(pop) > 2 else None |
|
81
|
|
|
r = rnd.choice(len(pop), 2, replace=not len(pop) >= 2, p=p) |
|
82
|
|
|
x = [x_b[i] + f * (pop[r[0]][i] - pop[r[1]][i]) if rnd.rand() < cr or i == j else pop[ic][i] for i in range(len(pop[ic]))] |
|
83
|
|
|
return asarray(x) |
|
84
|
|
|
|
|
85
|
|
View Code Duplication |
def CrossRand2(pop, ic, x_b, f, cr, rnd=rand, *args): |
|
|
|
|
|
|
86
|
|
|
r"""Mutation strategy with crossover. |
|
87
|
|
|
|
|
88
|
|
|
Mutation strategy uses five different random individuals from population. |
|
89
|
|
|
|
|
90
|
|
|
Mutation: |
|
91
|
|
|
Name: de/best/1 |
|
92
|
|
|
|
|
93
|
|
|
:math:`\mathbf{v}_{i, G} = \mathbf{x}_{r_1, G} + F \cdot (\mathbf{x}_{r_2, G} - \mathbf{x}_{r_3, G}) + F \cdot (\mathbf{x}_{r_4, G} - \mathbf{x}_{r_5, G})` |
|
94
|
|
|
where :math:`r_1, r_2, r_3, r_4, r_5` are random indexes representing current population individuals. |
|
95
|
|
|
|
|
96
|
|
|
Crossover: |
|
97
|
|
|
Name: Binomial crossover |
|
98
|
|
|
|
|
99
|
|
|
:math:`\mathbf{x}_{i, G+1} = \begin{cases} \mathbf{u}_{i, G+1}, & \text{if $f(\mathbf{u}_{i, G+1}) \leq f(\mathbf{x}_{i, G})$}, \\ \mathbf{x}_{i, G}, & \text{otherwise}. \end{cases}` |
|
100
|
|
|
|
|
101
|
|
|
Args: |
|
102
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
103
|
|
|
ic (int): Index of individual being mutated. |
|
104
|
|
|
x_b (Individual): Current global best individual. |
|
105
|
|
|
f (float): Scale factor. |
|
106
|
|
|
cr (float): Crossover probability. |
|
107
|
|
|
rnd (mtrand.RandomState): Random generator. |
|
108
|
|
|
*args (list): Additional arguments. |
|
109
|
|
|
|
|
110
|
|
|
Returns: |
|
111
|
|
|
numpy.ndarray: mutated and mixed individual. |
|
112
|
|
|
""" |
|
113
|
|
|
j = rnd.randint(len(pop[ic])) |
|
114
|
|
|
p = [1 / (len(pop) - 1.0) if i != ic else 0 for i in range(len(pop))] if len(pop) > 5 else None |
|
115
|
|
|
r = rnd.choice(len(pop), 5, replace=not len(pop) >= 5, p=p) |
|
116
|
|
|
x = [pop[r[0]][i] + f * (pop[r[1]][i] - pop[r[2]][i]) + f * (pop[r[3]][i] - pop[r[4]][i]) if rnd.rand() < cr or i == j else pop[ic][i] for i in range(len(pop[ic]))] |
|
117
|
|
|
return asarray(x) |
|
118
|
|
|
|
|
119
|
|
View Code Duplication |
def CrossBest2(pop, ic, x_b, f, cr, rnd=rand, *args): |
|
|
|
|
|
|
120
|
|
|
r"""Mutation strategy with crossover. |
|
121
|
|
|
|
|
122
|
|
|
Mutation: |
|
123
|
|
|
Name: de/best/2 |
|
124
|
|
|
|
|
125
|
|
|
:math:`\mathbf{v}_{i, G} = \mathbf{x}_{best, G} + F \cdot (\mathbf{x}_{r_1, G} - \mathbf{x}_{r_2, G}) + F \cdot (\mathbf{x}_{r_3, G} - \mathbf{x}_{r_4, G})` |
|
126
|
|
|
where :math:`r_1, r_2, r_3, r_4` are random indexes representing current population individuals. |
|
127
|
|
|
|
|
128
|
|
|
Crossover: |
|
129
|
|
|
Name: Binomial crossover |
|
130
|
|
|
|
|
131
|
|
|
:math:`\mathbf{x}_{i, G+1} = \begin{cases} \mathbf{u}_{i, G+1}, & \text{if $f(\mathbf{u}_{i, G+1}) \leq f(\mathbf{x}_{i, G})$}, \\ \mathbf{x}_{i, G}, & \text{otherwise}. \end{cases}` |
|
132
|
|
|
|
|
133
|
|
|
Args: |
|
134
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
135
|
|
|
ic (int): Index of individual being mutated. |
|
136
|
|
|
x_b (Individual): Current global best individual. |
|
137
|
|
|
f (float): Scale factor. |
|
138
|
|
|
cr (float): Crossover probability. |
|
139
|
|
|
rnd (mtrand.RandomState): Random generator. |
|
140
|
|
|
*args (list): Additional arguments. |
|
141
|
|
|
|
|
142
|
|
|
Returns: |
|
143
|
|
|
numpy.ndarray: mutated and mixed individual. |
|
144
|
|
|
""" |
|
145
|
|
|
j = rnd.randint(len(pop[ic])) |
|
146
|
|
|
p = [1 / (len(pop) - 1.0) if i != ic else 0 for i in range(len(pop))] if len(pop) > 4 else None |
|
147
|
|
|
r = rnd.choice(len(pop), 4, replace=not len(pop) >= 4, p=p) |
|
148
|
|
|
x = [x_b[i] + f * (pop[r[0]][i] - pop[r[1]][i]) + f * (pop[r[2]][i] - pop[r[3]][i]) if rnd.rand() < cr or i == j else pop[ic][i] for i in range(len(pop[ic]))] |
|
149
|
|
|
return asarray(x) |
|
150
|
|
|
|
|
151
|
|
View Code Duplication |
def CrossCurr2Rand1(pop, ic, x_b, f, cr, rnd=rand, *args): |
|
|
|
|
|
|
152
|
|
|
r"""Mutation strategy with crossover. |
|
153
|
|
|
|
|
154
|
|
|
Mutation: |
|
155
|
|
|
Name: de/curr2rand/1 |
|
156
|
|
|
|
|
157
|
|
|
:math:`\mathbf{v}_{i, G} = \mathbf{x}_{i, G} + F \cdot (\mathbf{x}_{r_1, G} - \mathbf{x}_{r_2, G}) + F \cdot (\mathbf{x}_{r_3, G} - \mathbf{x}_{r_4, G})` |
|
158
|
|
|
where :math:`r_1, r_2, r_3, r_4` are random indexes representing current population individuals |
|
159
|
|
|
|
|
160
|
|
|
Crossover: |
|
161
|
|
|
Name: Binomial crossover |
|
162
|
|
|
|
|
163
|
|
|
:math:`\mathbf{x}_{i, G+1} = \begin{cases} \mathbf{u}_{i, G+1}, & \text{if $f(\mathbf{u}_{i, G+1}) \leq f(\mathbf{x}_{i, G})$}, \\ \mathbf{x}_{i, G}, & \text{otherwise}. \end{cases}` |
|
164
|
|
|
|
|
165
|
|
|
Args: |
|
166
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
167
|
|
|
ic (int): Index of individual being mutated. |
|
168
|
|
|
x_b (Individual): Current global best individual. |
|
169
|
|
|
f (float): Scale factor. |
|
170
|
|
|
cr (float): Crossover probability. |
|
171
|
|
|
rnd (mtrand.RandomState): Random generator. |
|
172
|
|
|
*args (list): Additional arguments. |
|
173
|
|
|
|
|
174
|
|
|
Returns: |
|
175
|
|
|
numpy.ndarray: mutated and mixed individual. |
|
176
|
|
|
""" |
|
177
|
|
|
j = rnd.randint(len(pop[ic])) |
|
178
|
|
|
p = [1 / (len(pop) - 1.0) if i != ic else 0 for i in range(len(pop))] if len(pop) > 4 else None |
|
179
|
|
|
r = rnd.choice(len(pop), 4, replace=not len(pop) >= 4, p=p) |
|
180
|
|
|
x = [pop[ic][i] + f * (pop[r[0]][i] - pop[r[1]][i]) + f * (pop[r[2]][i] - pop[r[3]][i]) if rnd.rand() < cr or i == j else pop[ic][i] for i in range(len(pop[ic]))] |
|
181
|
|
|
return asarray(x) |
|
182
|
|
|
|
|
183
|
|
|
def CrossCurr2Best1(pop, ic, x_b, f, cr, rnd=rand, **kwargs): |
|
184
|
|
|
r"""Mutation strategy with crossover. |
|
185
|
|
|
|
|
186
|
|
|
Mutation: |
|
187
|
|
|
Name: de/curr-to-best/1 |
|
188
|
|
|
|
|
189
|
|
|
:math:`\mathbf{v}_{i, G} = \mathbf{x}_{i, G} + F \cdot (\mathbf{x}_{r_1, G} - \mathbf{x}_{r_2, G}) + F \cdot (\mathbf{x}_{r_3, G} - \mathbf{x}_{r_4, G})` |
|
190
|
|
|
where :math:`r_1, r_2, r_3, r_4` are random indexes representing current population individuals |
|
191
|
|
|
|
|
192
|
|
|
Crossover: |
|
193
|
|
|
Name: Binomial crossover |
|
194
|
|
|
|
|
195
|
|
|
:math:`\mathbf{x}_{i, G+1} = \begin{cases} \mathbf{u}_{i, G+1}, & \text{if $f(\mathbf{u}_{i, G+1}) \leq f(\mathbf{x}_{i, G})$}, \\ \mathbf{x}_{i, G}, & \text{otherwise}. \end{cases}` |
|
196
|
|
|
|
|
197
|
|
|
Args: |
|
198
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
199
|
|
|
ic (int): Index of individual being mutated. |
|
200
|
|
|
x_b (Individual): Current global best individual. |
|
201
|
|
|
f (float): Scale factor. |
|
202
|
|
|
cr (float): Crossover probability. |
|
203
|
|
|
rnd (mtrand.RandomState): Random generator. |
|
204
|
|
|
*args (list): Additional arguments. |
|
205
|
|
|
|
|
206
|
|
|
Returns: |
|
207
|
|
|
numpy.ndarray: mutated and mixed individual. |
|
208
|
|
|
""" |
|
209
|
|
|
j = rnd.randint(len(pop[ic])) |
|
210
|
|
|
p = [1 / (len(pop) - 1.0) if i != ic else 0 for i in range(len(pop))] if len(pop) > 3 else None |
|
211
|
|
|
r = rnd.choice(len(pop), 3, replace=not len(pop) >= 3, p=p) |
|
212
|
|
|
x = [pop[ic][i] + f * (x_b[i] - pop[r[0]][i]) + f * (pop[r[1]][i] - pop[r[2]][i]) if rnd.rand() < cr or i == j else pop[ic][i] for i in range(len(pop[ic]))] |
|
213
|
|
|
return asarray(x) |
|
214
|
|
|
|
|
215
|
|
|
class DifferentialEvolution(Algorithm): |
|
216
|
|
|
r"""Implementation of Differential evolution algorithm. |
|
217
|
|
|
|
|
218
|
|
|
Algorithm: |
|
219
|
|
|
Differential evolution algorithm |
|
220
|
|
|
|
|
221
|
|
|
Date: |
|
222
|
|
|
2018 |
|
223
|
|
|
|
|
224
|
|
|
Author: |
|
225
|
|
|
Uros Mlakar and Klemen Berkovič |
|
226
|
|
|
|
|
227
|
|
|
License: |
|
228
|
|
|
MIT |
|
229
|
|
|
|
|
230
|
|
|
Reference paper: |
|
231
|
|
|
Storn, Rainer, and Kenneth Price. "Differential evolution - a simple and efficient heuristic for global optimization over continuous spaces." Journal of global optimization 11.4 (1997): 341-359. |
|
232
|
|
|
|
|
233
|
|
|
Attributes: |
|
234
|
|
|
Name (List[str]): List of string of names for algorithm. |
|
235
|
|
|
F (float): Scale factor. |
|
236
|
|
|
CR (float): Crossover probability. |
|
237
|
|
|
CrossMutt (Callable[numpy.ndarray, int, numpy.ndarray, float, float, mtrand.RandomState, Dict[str, Any]]): crossover and mutation strategy. |
|
238
|
|
|
|
|
239
|
|
|
See Also: |
|
240
|
|
|
* :class:`NiaPy.algorithms.Algorithm` |
|
241
|
|
|
""" |
|
242
|
|
|
Name = ['DifferentialEvolution', 'DE'] |
|
243
|
|
|
|
|
244
|
|
|
@staticmethod |
|
245
|
|
|
def algorithmInfo(): |
|
246
|
|
|
r"""Get basic information of algorithm. |
|
247
|
|
|
|
|
248
|
|
|
Returns: |
|
249
|
|
|
str: Basic information of algorithm. |
|
250
|
|
|
|
|
251
|
|
|
See Also: |
|
252
|
|
|
* :func:`NiaPy.algorithms.Algorithm.algorithmInfo` |
|
253
|
|
|
""" |
|
254
|
|
|
return r"""Storn, Rainer, and Kenneth Price. "Differential evolution - a simple and efficient heuristic for global optimization over continuous spaces." Journal of global optimization 11.4 (1997): 341-359.""" |
|
255
|
|
|
|
|
256
|
|
|
@staticmethod |
|
257
|
|
|
def typeParameters(): |
|
258
|
|
|
r"""Get dictionary with functions for checking values of parameters. |
|
259
|
|
|
|
|
260
|
|
|
Returns: |
|
261
|
|
|
Dict[str, Callable]: |
|
262
|
|
|
* F (Callable[[Union[float, int]], bool]): Check for correct value of parameter. |
|
263
|
|
|
* CR (Callable[[float], bool]): Check for correct value of parameter. |
|
264
|
|
|
|
|
265
|
|
|
See Also: |
|
266
|
|
|
* :func:`NiaPy.algorithms.Algorithm.typeParameters` |
|
267
|
|
|
""" |
|
268
|
|
|
d = Algorithm.typeParameters() |
|
269
|
|
|
d.update({ |
|
270
|
|
|
'F': lambda x: isinstance(x, (float, int)) and 0 < x <= 2, |
|
271
|
|
|
'CR': lambda x: isinstance(x, float) and 0 <= x <= 1 |
|
272
|
|
|
}) |
|
273
|
|
|
return d |
|
274
|
|
|
|
|
275
|
|
|
def setParameters(self, NP=50, F=1, CR=0.8, CrossMutt=CrossRand1, **ukwargs): |
|
276
|
|
|
r"""Set the algorithm parameters. |
|
277
|
|
|
|
|
278
|
|
|
Arguments: |
|
279
|
|
|
NP (Optional[int]): Population size. |
|
280
|
|
|
F (Optional[float]): Scaling factor. |
|
281
|
|
|
CR (Optional[float]): Crossover rate. |
|
282
|
|
|
CrossMutt (Optional[Callable[[numpy.ndarray, int, numpy.ndarray, float, float, mtrand.RandomState, list], numpy.ndarray]]): Crossover and mutation strategy. |
|
283
|
|
|
ukwargs (Dict[str, Any]): Additional arguments. |
|
284
|
|
|
|
|
285
|
|
|
See Also: |
|
286
|
|
|
* :func:`NiaPy.algorithms.Algorithm.setParameters` |
|
287
|
|
|
""" |
|
288
|
|
|
Algorithm.setParameters(self, NP=NP, InitPopFunc=ukwargs.pop('InitPopFunc', defaultIndividualInit), itype=ukwargs.pop('itype', Individual), **ukwargs) |
|
289
|
|
|
self.F, self.CR, self.CrossMutt = F, CR, CrossMutt |
|
290
|
|
|
if ukwargs: logger.info('Unused arguments: %s' % (ukwargs)) |
|
291
|
|
|
|
|
292
|
|
|
def evolve(self, pop, xb, task, **kwargs): |
|
293
|
|
|
r"""Evolve population. |
|
294
|
|
|
|
|
295
|
|
|
Args: |
|
296
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
297
|
|
|
xb (Individual): Current best individual. |
|
298
|
|
|
task (Task): Optimization task. |
|
299
|
|
|
**kwargs (Dict[str, Any]): Additional arguments. |
|
300
|
|
|
|
|
301
|
|
|
Returns: |
|
302
|
|
|
numpy.ndarray[Individual]: New evolved populations. |
|
303
|
|
|
""" |
|
304
|
|
|
return objects2array([self.itype(x=self.CrossMutt(pop, i, xb, self.F, self.CR, self.Rand), task=task, rnd=self.Rand, e=True) for i in range(len(pop))]) |
|
305
|
|
|
|
|
306
|
|
|
def selection(self, pop, npop, **kwargs): |
|
307
|
|
|
r"""Operator for selection. |
|
308
|
|
|
|
|
309
|
|
|
Args: |
|
310
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
311
|
|
|
npop (numpy.ndarray[Individual]): New Population. |
|
312
|
|
|
**kwargs (Dict[str, Any]): Additional arguments. |
|
313
|
|
|
|
|
314
|
|
|
Returns: |
|
315
|
|
|
numpy.ndarray[Individual]: New selected individuals. |
|
316
|
|
|
""" |
|
317
|
|
|
return objects2array([e if e.f < pop[i].f else pop[i] for i, e in enumerate(npop)]) |
|
318
|
|
|
|
|
319
|
|
|
def postSelection(self, pop, task, xb=None, **kwargs): |
|
320
|
|
|
r"""Apply additional operation after selection. |
|
321
|
|
|
|
|
322
|
|
|
Args: |
|
323
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
324
|
|
|
task (Task): Optimization task. |
|
325
|
|
|
xb (Optional[Individual]): Global best solution. |
|
326
|
|
|
**kwargs (Dict[str, Any]): Additional arguments. |
|
327
|
|
|
|
|
328
|
|
|
Returns: |
|
329
|
|
|
numpy.ndarray[Individual]: New population. |
|
330
|
|
|
""" |
|
331
|
|
|
return pop |
|
332
|
|
|
|
|
333
|
|
|
def runIteration(self, task, pop, fpop, xb, fxb, **dparams): |
|
334
|
|
|
r"""Core function of Differential Evolution algorithm. |
|
335
|
|
|
|
|
336
|
|
|
Args: |
|
337
|
|
|
task (Task): Optimization task. |
|
338
|
|
|
pop (numpy.ndarray[Initialized]): Current population. |
|
339
|
|
|
fpop (numpy.ndarray[float]): Current populations fitness/function values. |
|
340
|
|
|
xb (Individual): Current best individual. |
|
341
|
|
|
fxb (float): Current best individual function/fitness value. |
|
342
|
|
|
**dparams (Dict[str, Any]): Additional arguments. |
|
343
|
|
|
|
|
344
|
|
|
Returns: |
|
345
|
|
|
Tuple[numpy.ndarray[Individual], numpy.ndarray[float], Dict[str, Any]]: |
|
346
|
|
|
1. New population. |
|
347
|
|
|
2. New population fitness/function values. |
|
348
|
|
|
3. Additional arguments. |
|
349
|
|
|
|
|
350
|
|
|
See Also: |
|
351
|
|
|
* :func:`NiaPy.algorithms.basic.DifferentialEvolution.evolve` |
|
352
|
|
|
* :func:`NiaPy.algorithms.basic.DifferentialEvolution.selection` |
|
353
|
|
|
* :func:`NiaPy.algorithms.basic.DifferentialEvolution.postSelection` |
|
354
|
|
|
""" |
|
355
|
|
|
npop = self.evolve(pop, xb, task) |
|
356
|
|
|
pop = self.selection(pop, npop, task=task) |
|
357
|
|
|
pop = self.postSelection(pop, task, xb=xb) |
|
358
|
|
|
return pop, asarray([x.f for x in pop]), {} |
|
359
|
|
|
|
|
360
|
|
|
class CrowdingDifferentialEvolution(DifferentialEvolution): |
|
361
|
|
|
r"""Implementation of Differential evolution algorithm with multiple mutation strateys. |
|
362
|
|
|
|
|
363
|
|
|
Algorithm: |
|
364
|
|
|
Implementation of Differential evolution algorithm with multiple mutation strateys |
|
365
|
|
|
|
|
366
|
|
|
Date: |
|
367
|
|
|
2018 |
|
368
|
|
|
|
|
369
|
|
|
Author: |
|
370
|
|
|
Klemen Berkovič |
|
371
|
|
|
|
|
372
|
|
|
License: |
|
373
|
|
|
MIT |
|
374
|
|
|
|
|
375
|
|
|
Attributes: |
|
376
|
|
|
Name (List[str]): List of strings representing algorithm name. |
|
377
|
|
|
CrowPop (float): Proportion of range for cowding. |
|
378
|
|
|
|
|
379
|
|
|
See Also: |
|
380
|
|
|
* :class:`NiaPy.algorithms.basic.DifferentialEvolution` |
|
381
|
|
|
""" |
|
382
|
|
|
Name = ['CrowdingDifferentialEvolution', 'CDE'] |
|
383
|
|
|
|
|
384
|
|
|
@staticmethod |
|
385
|
|
|
def algorithmInfo(): |
|
386
|
|
|
r"""Get basic information of algorithm. |
|
387
|
|
|
|
|
388
|
|
|
Returns: |
|
389
|
|
|
str: Basic information of algorithm. |
|
390
|
|
|
|
|
391
|
|
|
See Also: |
|
392
|
|
|
* :func:`NiaPy.algorithms.Algorithm.algorithmInfo` |
|
393
|
|
|
""" |
|
394
|
|
|
return r"""No New""" |
|
395
|
|
|
|
|
396
|
|
|
def setParameters(self, CrowPop=0.1, **ukwargs): |
|
397
|
|
|
r"""Set core parameters of algorithm. |
|
398
|
|
|
|
|
399
|
|
|
Args: |
|
400
|
|
|
CrowPop (Optional[float]): Crowding distance. |
|
401
|
|
|
**ukwargs: Additional arguments. |
|
402
|
|
|
|
|
403
|
|
|
See Also: |
|
404
|
|
|
* :func:`NiaPy.algorithms.basic.DifferentialEvolution.setParameters` |
|
405
|
|
|
""" |
|
406
|
|
|
DifferentialEvolution.setParameters(self, **ukwargs) |
|
407
|
|
|
self.CrowPop = CrowPop |
|
408
|
|
|
|
|
409
|
|
|
def selection(self, pop, npop, **kwargs): |
|
410
|
|
|
r"""Operator for selection of individuals. |
|
411
|
|
|
|
|
412
|
|
|
Args: |
|
413
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
414
|
|
|
npop (numpy.ndarray[Individual]): New population. |
|
415
|
|
|
kwargs (Dict[str, Any]): Additional arguments. |
|
416
|
|
|
|
|
417
|
|
|
Returns: |
|
418
|
|
|
numpy.ndarray[Individual]: New population. |
|
419
|
|
|
""" |
|
420
|
|
|
P = [] |
|
421
|
|
|
for e in npop: |
|
422
|
|
|
i = argmin([euclidean(e, f) for f in pop]) |
|
423
|
|
|
P.append(pop[i] if pop[i].f < e.f else e) |
|
424
|
|
|
return asarray(P) |
|
425
|
|
|
|
|
426
|
|
|
class DynNpDifferentialEvolution(DifferentialEvolution): |
|
427
|
|
|
r"""Implementation of Dynamic poulation size Differential evolution algorithm. |
|
428
|
|
|
|
|
429
|
|
|
Algorithm: |
|
430
|
|
|
Dynamic poulation size Differential evolution algorithm |
|
431
|
|
|
|
|
432
|
|
|
Date: |
|
433
|
|
|
2018 |
|
434
|
|
|
|
|
435
|
|
|
Author: |
|
436
|
|
|
Klemen Berkovič |
|
437
|
|
|
|
|
438
|
|
|
License: |
|
439
|
|
|
MIT |
|
440
|
|
|
|
|
441
|
|
|
Attributes: |
|
442
|
|
|
Name (List[str]): List of strings representing algorithm names. |
|
443
|
|
|
pmax (int): Number of population reductions. |
|
444
|
|
|
rp (int): Small non-negative number which is added to value of generations. |
|
445
|
|
|
|
|
446
|
|
|
See Also: |
|
447
|
|
|
* :class:`NiaPy.algorithms.basic.DifferentialEvolution` |
|
448
|
|
|
""" |
|
449
|
|
|
Name = ['DynNpDifferentialEvolution', 'dynNpDE'] |
|
450
|
|
|
|
|
451
|
|
|
@staticmethod |
|
452
|
|
|
def algorithmInfo(): |
|
453
|
|
|
r"""Get basic information of algorithm. |
|
454
|
|
|
|
|
455
|
|
|
Returns: |
|
456
|
|
|
str: Basic information of algorithm. |
|
457
|
|
|
|
|
458
|
|
|
See Also: |
|
459
|
|
|
* :func:`NiaPy.algorithms.Algorithm.algorithmInfo` |
|
460
|
|
|
""" |
|
461
|
|
|
return r"""No info""" |
|
462
|
|
|
|
|
463
|
|
|
@staticmethod |
|
464
|
|
|
def typeParameters(): |
|
465
|
|
|
r"""Get dictionary with functions for checking values of parameters. |
|
466
|
|
|
|
|
467
|
|
|
Returns: |
|
468
|
|
|
Dict[str, Callable]: |
|
469
|
|
|
* rp (Callable[[Union[float, int]], bool]) |
|
470
|
|
|
* pmax (Callable[[int], bool]) |
|
471
|
|
|
|
|
472
|
|
|
See Also: |
|
473
|
|
|
* :func:`NiaPy.algorithms.basic.DifferentialEvolution.typeParameters` |
|
474
|
|
|
""" |
|
475
|
|
|
r = DifferentialEvolution.typeParameters() |
|
476
|
|
|
r['rp'] = lambda x: isinstance(x, (float, int)) and x > 0 |
|
477
|
|
|
r['pmax'] = lambda x: isinstance(x, int) and x > 0 |
|
478
|
|
|
return r |
|
479
|
|
|
|
|
480
|
|
|
def setParameters(self, pmax=50, rp=3, **ukwargs): |
|
481
|
|
|
r"""Set the algorithm parameters. |
|
482
|
|
|
|
|
483
|
|
|
Arguments: |
|
484
|
|
|
pmax (Optional[int]): umber of population reductions. |
|
485
|
|
|
rp (Optional[int]): Small non-negative number which is added to value of generations. |
|
486
|
|
|
|
|
487
|
|
|
See Also: |
|
488
|
|
|
* :func:`NiaPy.algorithms.basic.DifferentialEvolution.setParameters` |
|
489
|
|
|
""" |
|
490
|
|
|
DifferentialEvolution.setParameters(self, **ukwargs) |
|
491
|
|
|
self.pmax, self.rp = pmax, rp |
|
492
|
|
|
if ukwargs: logger.info('Unused arguments: %s' % (ukwargs)) |
|
493
|
|
|
|
|
494
|
|
|
def postSelection(self, pop, task, **kwargs): |
|
495
|
|
|
r"""Post selection operator. |
|
496
|
|
|
|
|
497
|
|
|
In this algorithm the post selection operator decrements the population at specific iterations/generations. |
|
498
|
|
|
|
|
499
|
|
|
Args: |
|
500
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
501
|
|
|
task (Task): Optimization task. |
|
502
|
|
|
kwargs (Dict[str, Any]): Additional arguments. |
|
503
|
|
|
|
|
504
|
|
|
Returns: |
|
505
|
|
|
numpy.ndarray[Individual]: Changed current population. |
|
506
|
|
|
""" |
|
507
|
|
|
Gr = task.nFES // (self.pmax * len(pop)) + self.rp |
|
508
|
|
|
nNP = len(pop) // 2 |
|
509
|
|
|
if task.Iters == Gr and len(pop) > 3: pop = objects2array([pop[i] if pop[i].f < pop[i + nNP].f else pop[i + nNP] for i in range(nNP)]) |
|
510
|
|
|
return pop |
|
511
|
|
|
|
|
512
|
|
|
def proportional(Lt_min, Lt_max, mu, x_f, avg, *args): |
|
513
|
|
|
r"""Proportional calculation of age of individual. |
|
514
|
|
|
|
|
515
|
|
|
Args: |
|
516
|
|
|
Lt_min (int): Minimal life time. |
|
517
|
|
|
Lt_max (int): Maximal life time. |
|
518
|
|
|
mu (float): Median of life time. |
|
519
|
|
|
x_f (float): Individuals function/fitness value. |
|
520
|
|
|
avg (float): Average fitness/function value of current population. |
|
521
|
|
|
*args (list): Additional arguments. |
|
522
|
|
|
|
|
523
|
|
|
Returns: |
|
524
|
|
|
int: Age of individual. |
|
525
|
|
|
""" |
|
526
|
|
|
return min(Lt_min + mu * avg / x_f, Lt_max) |
|
527
|
|
|
|
|
528
|
|
|
def linear(Lt_min, Lt_max, mu, x_f, avg, x_gw, x_gb, *args): |
|
529
|
|
|
r"""Linear calculation of age of individual. |
|
530
|
|
|
|
|
531
|
|
|
Args: |
|
532
|
|
|
Lt_min (int): Minimal life time. |
|
533
|
|
|
Lt_max (int): Maximal life time. |
|
534
|
|
|
mu (float): Median of life time. |
|
535
|
|
|
x_f (float): Individual function/fitness value. |
|
536
|
|
|
avg (float): Average fitness/function value. |
|
537
|
|
|
x_gw (float): Global worst fitness/function value. |
|
538
|
|
|
x_gb (float): Global best fitness/function value. |
|
539
|
|
|
*args (list): Additional arguments. |
|
540
|
|
|
|
|
541
|
|
|
Returns: |
|
542
|
|
|
int: Age of individual. |
|
543
|
|
|
""" |
|
544
|
|
|
return Lt_min + 2 * mu * (x_f - x_gw) / (x_gb - x_gw) |
|
545
|
|
|
|
|
546
|
|
|
def bilinear(Lt_min, Lt_max, mu, x_f, avg, x_gw, x_gb, *args): |
|
547
|
|
|
r"""Bilinear calculation of age of individual. |
|
548
|
|
|
|
|
549
|
|
|
Args: |
|
550
|
|
|
Lt_min (int): Minimal life time. |
|
551
|
|
|
Lt_max (int): Maximal life time. |
|
552
|
|
|
mu (float): Median of life time. |
|
553
|
|
|
x_f (float): Individual function/fitness value. |
|
554
|
|
|
avg (float): Average fitness/function value. |
|
555
|
|
|
x_gw (float): Global worst fitness/function value. |
|
556
|
|
|
x_gb (float): Global best fitness/function value. |
|
557
|
|
|
*args (list): Additional arguments. |
|
558
|
|
|
|
|
559
|
|
|
Returns: |
|
560
|
|
|
int: Age of individual. |
|
561
|
|
|
""" |
|
562
|
|
|
if avg < x_f: return Lt_min + mu * (x_f - x_gw) / (x_gb - x_gw) |
|
563
|
|
|
return 0.5 * (Lt_min + Lt_max) + mu * (x_f - avg) / (x_gb - avg) |
|
564
|
|
|
|
|
565
|
|
|
class AgingIndividual(Individual): |
|
566
|
|
|
r"""Individual with aging. |
|
567
|
|
|
|
|
568
|
|
|
Attributes: |
|
569
|
|
|
age (int): Age of individual. |
|
570
|
|
|
|
|
571
|
|
|
See Also: |
|
572
|
|
|
* :class:`NiaPy.algorithms.Individual` |
|
573
|
|
|
""" |
|
574
|
|
|
age = 0 |
|
575
|
|
|
|
|
576
|
|
|
def __init__(self, **kwargs): |
|
577
|
|
|
r"""Init Aging Individual. |
|
578
|
|
|
|
|
579
|
|
|
Args: |
|
580
|
|
|
**kwargs (Dict[str, Any]): Additional arguments sent to parent. |
|
581
|
|
|
|
|
582
|
|
|
See Also: |
|
583
|
|
|
* :func:`NiaPy.algorithms.Individual.__init__` |
|
584
|
|
|
""" |
|
585
|
|
|
Individual.__init__(self, **kwargs) |
|
586
|
|
|
self.age = 0 |
|
587
|
|
|
|
|
588
|
|
|
class AgingNpDifferentialEvolution(DifferentialEvolution): |
|
589
|
|
|
r"""Implementation of Differential evolution algorithm with aging individuals. |
|
590
|
|
|
|
|
591
|
|
|
Algorithm: |
|
592
|
|
|
Differential evolution algorithm with dynamic population size that is defined by the quality of population |
|
593
|
|
|
|
|
594
|
|
|
Date: |
|
595
|
|
|
2018 |
|
596
|
|
|
|
|
597
|
|
|
Author: |
|
598
|
|
|
Klemen Berkovič |
|
599
|
|
|
|
|
600
|
|
|
License: |
|
601
|
|
|
MIT |
|
602
|
|
|
|
|
603
|
|
|
Attributes: |
|
604
|
|
|
Name (List[str]): list of strings representing algorithm names. |
|
605
|
|
|
Lt_min (int): Minimal age of individual. |
|
606
|
|
|
Lt_max (int): Maximal age of individual. |
|
607
|
|
|
delta_np (float): Proportion of how many individuals shall die. |
|
608
|
|
|
omega (float): Acceptance rate for individuals to die. |
|
609
|
|
|
mu (int): Mean of individual max and min age. |
|
610
|
|
|
age (Callable[[int, int, float, float, float, float, float], int]): Function for calculation of age for individual. |
|
611
|
|
|
|
|
612
|
|
|
See Also: |
|
613
|
|
|
* :class:`NiaPy.algorithms.basic.DifferentialEvolution` |
|
614
|
|
|
""" |
|
615
|
|
|
Name = ['AgingNpDifferentialEvolution', 'ANpDE'] |
|
616
|
|
|
|
|
617
|
|
|
@staticmethod |
|
618
|
|
|
def algorithmInfo(): |
|
619
|
|
|
r"""Get basic information of algorithm. |
|
620
|
|
|
|
|
621
|
|
|
Returns: |
|
622
|
|
|
str: Basic information of algorithm. |
|
623
|
|
|
|
|
624
|
|
|
See Also: |
|
625
|
|
|
* :func:`NiaPy.algorithms.Algorithm.algorithmInfo` |
|
626
|
|
|
""" |
|
627
|
|
|
return r"""No info""" |
|
628
|
|
|
|
|
629
|
|
|
@staticmethod |
|
630
|
|
|
def typeParameters(): |
|
631
|
|
|
r"""Get dictionary with functions for checking values of parameters. |
|
632
|
|
|
|
|
633
|
|
|
Returns: |
|
634
|
|
|
Dict[str, Callable]: |
|
635
|
|
|
* Lt_min (Callable[[int], bool]) |
|
636
|
|
|
* Lt_max (Callable[[int], bool]) |
|
637
|
|
|
* delta_np (Callable[[float], bool]) |
|
638
|
|
|
* omega (Callable[[float], bool]) |
|
639
|
|
|
|
|
640
|
|
|
See Also: |
|
641
|
|
|
* :func:`NiaPy.algorithms.basic.DifferentialEvolution.typeParameters` |
|
642
|
|
|
""" |
|
643
|
|
|
r = DifferentialEvolution.typeParameters() |
|
644
|
|
|
r.update({ |
|
645
|
|
|
'Lt_min': lambda x: isinstance(x, int) and x >= 0, |
|
646
|
|
|
'Lt_max': lambda x: isinstance(x, int) and x >= 0, |
|
647
|
|
|
'delta_np': lambda x: isinstance(x, float) and 0 <= x <= 1, |
|
648
|
|
|
'omega': lambda x: isinstance(x, float) and 1 >= x >= 0 |
|
649
|
|
|
}) |
|
650
|
|
|
return r |
|
651
|
|
|
|
|
652
|
|
|
def setParameters(self, Lt_min=0, Lt_max=12, delta_np=0.3, omega=0.3, age=proportional, CrossMutt=CrossBest1, **ukwargs): |
|
653
|
|
|
r"""Set the algorithm parameters. |
|
654
|
|
|
|
|
655
|
|
|
Arguments: |
|
656
|
|
|
Lt_min (Optional[int]): Minimum life time. |
|
657
|
|
|
Lt_max (Optional[int]): Maximum life time. |
|
658
|
|
|
age (Optional[Callable[[int, int, float, float, float, float, float], int]]): Function for calculation of age for individual. |
|
659
|
|
|
|
|
660
|
|
|
See Also: |
|
661
|
|
|
* :func:`NiaPy.algorithms.basic.DifferentialEvolution.setParameters` |
|
662
|
|
|
""" |
|
663
|
|
|
DifferentialEvolution.setParameters(self, itype=AgingIndividual, **ukwargs) |
|
664
|
|
|
self.Lt_min, self.Lt_max, self.age, self.delta_np, self.omega = Lt_min, Lt_max, age, delta_np, omega |
|
665
|
|
|
self.mu = abs(self.Lt_max - self.Lt_min) / 2 |
|
666
|
|
|
if ukwargs: logger.info('Unused arguments: %s' % (ukwargs)) |
|
667
|
|
|
|
|
668
|
|
|
def deltaPopE(self, t): |
|
669
|
|
|
r"""Calculate how many individuals are going to dye. |
|
670
|
|
|
|
|
671
|
|
|
Args: |
|
672
|
|
|
t (int): Number of generations made by the algorithm. |
|
673
|
|
|
|
|
674
|
|
|
Returns: |
|
675
|
|
|
int: Number of individuals to dye. |
|
676
|
|
|
""" |
|
677
|
|
|
return int(self.delta_np * abs(sin(t))) |
|
678
|
|
|
|
|
679
|
|
|
def deltaPopC(self, t): |
|
680
|
|
|
r"""Calculate how many individuals are going to be created. |
|
681
|
|
|
|
|
682
|
|
|
Args: |
|
683
|
|
|
t (int): Number of generations made by the algorithm. |
|
684
|
|
|
|
|
685
|
|
|
Returns: |
|
686
|
|
|
int: Number of individuals to be born. |
|
687
|
|
|
""" |
|
688
|
|
|
return int(self.delta_np * abs(cos(t))) |
|
689
|
|
|
|
|
690
|
|
|
def aging(self, task, pop): |
|
691
|
|
|
r"""Apply aging to individuals. |
|
692
|
|
|
|
|
693
|
|
|
Args: |
|
694
|
|
|
task (Task): Optimization task. |
|
695
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
696
|
|
|
|
|
697
|
|
|
Returns: |
|
698
|
|
|
numpy.ndarray[Individual]: New population. |
|
699
|
|
|
""" |
|
700
|
|
|
fpop = asarray([x.f for x in pop]) |
|
701
|
|
|
x_b, x_w = pop[argmin(fpop)], pop[argmax(fpop)] |
|
702
|
|
|
avg, npop = mean(fpop), [] |
|
703
|
|
|
for x in pop: |
|
704
|
|
|
x.age += 1 |
|
705
|
|
|
Lt = round(self.age(self.Lt_min, self.Lt_max, self.mu, x.f, avg, x_w, x_b)) |
|
706
|
|
|
if x.age <= Lt: npop.append(x) |
|
707
|
|
|
if len(npop) != 0: npop = objects2array([self.itype(task=task, rnd=self.Rand, e=True) for _i in range(len(pop))]) |
|
708
|
|
|
return npop |
|
709
|
|
|
|
|
710
|
|
|
def popIncrement(self, pop, task): |
|
711
|
|
|
r"""Increment population. |
|
712
|
|
|
|
|
713
|
|
|
Args: |
|
714
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
715
|
|
|
task (Task): Optimization task. |
|
716
|
|
|
|
|
717
|
|
|
Returns: |
|
718
|
|
|
numpy.ndarray[Individual]: Increased population. |
|
719
|
|
|
""" |
|
720
|
|
|
deltapop = int(round(max(1, self.NP * self.deltaPopE(task.Iters)))) |
|
721
|
|
|
return objects2array([self.itype(task=task, rnd=self.Rand, e=True) for _ in range(deltapop)]) |
|
722
|
|
|
|
|
723
|
|
|
def popDecrement(self, pop, task): |
|
724
|
|
|
r"""Decrement population. |
|
725
|
|
|
|
|
726
|
|
|
Args: |
|
727
|
|
|
pop (numpy.ndarray): Current population. |
|
728
|
|
|
task (Task): Optimization task. |
|
729
|
|
|
|
|
730
|
|
|
Returns: |
|
731
|
|
|
numpy.ndarray[Individual]: Decreased population. |
|
732
|
|
|
""" |
|
733
|
|
|
deltapop = int(round(max(1, self.NP * self.deltaPopC(task.Iters)))) |
|
734
|
|
|
if len(pop) - deltapop <= 0: return pop |
|
735
|
|
|
ni = self.Rand.choice(len(pop), deltapop, replace=False) |
|
736
|
|
|
npop = [] |
|
737
|
|
|
for i, e in enumerate(pop): |
|
738
|
|
|
if i not in ni: npop.append(e) |
|
739
|
|
|
elif self.rand() >= self.omega: npop.append(e) |
|
740
|
|
|
return objects2array(npop) |
|
741
|
|
|
|
|
742
|
|
|
def selection(self, pop, npop, task, **kwargs): |
|
743
|
|
|
r"""Select operator for individuals with aging. |
|
744
|
|
|
|
|
745
|
|
|
Args: |
|
746
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
747
|
|
|
npop (numpy.ndarray[Individual]): New population. |
|
748
|
|
|
task (Task): Optimization task. |
|
749
|
|
|
**kwargs (Dict[str, Any]): Additional arguments. |
|
750
|
|
|
|
|
751
|
|
|
Returns: |
|
752
|
|
|
numpy.ndarray[Individual]: New population of individuals. |
|
753
|
|
|
""" |
|
754
|
|
|
npop = DifferentialEvolution.selection(self, pop, npop) |
|
755
|
|
|
npop = append(npop, self.popIncrement(pop, task)) |
|
756
|
|
|
pop = self.aging(task, npop) |
|
757
|
|
|
return pop |
|
758
|
|
|
|
|
759
|
|
|
def postSelection(self, pop, task, xb=None, **kwargs): |
|
760
|
|
|
r"""Post selection operator. |
|
761
|
|
|
|
|
762
|
|
|
Args: |
|
763
|
|
|
pop (numpy.ndarray): Current population. |
|
764
|
|
|
task (Task): Optimization task. |
|
765
|
|
|
xb (Individual): Global best individual. |
|
766
|
|
|
**kwargs (Dict[str, Any]): Additional arguments. |
|
767
|
|
|
|
|
768
|
|
|
Returns: |
|
769
|
|
|
numpy.ndarray[Individual]: New population. |
|
770
|
|
|
""" |
|
771
|
|
|
return self.popDecrement(pop, task) if len(pop) > self.NP else pop |
|
772
|
|
|
|
|
773
|
|
|
def multiMutations(pop, i, xb, F, CR, rnd, task, itype, strategies, **kwargs): |
|
774
|
|
|
r"""Mutation strategy that takes more than one strategy and applys them to individual. |
|
775
|
|
|
|
|
776
|
|
|
Args: |
|
777
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
778
|
|
|
i (int): Index of current individual. |
|
779
|
|
|
xb (Individual): Current best individual. |
|
780
|
|
|
F (float): Scale factor. |
|
781
|
|
|
CR (float): Crossover probability. |
|
782
|
|
|
rnd (mtrand.RandomState): Random generator. |
|
783
|
|
|
task (Task): Optimization task. |
|
784
|
|
|
IndividualType (Individual): Individual type used in algorithm. |
|
785
|
|
|
strategies (Iterable[Callable[[numpy.ndarray[Individual], int, Individual, float, float, mtrand.RandomState], numpy.ndarray[Individual]]]): List of mutation strategies. |
|
786
|
|
|
**kwargs (Dict[str, Any]): Additional arguments. |
|
787
|
|
|
|
|
788
|
|
|
Returns: |
|
789
|
|
|
Individual: Best individual from applyed mutations strategies. |
|
790
|
|
|
""" |
|
791
|
|
|
L = [itype(x=strategy(pop, i, xb, F, CR, rnd=rnd), task=task, e=True, rnd=rnd) for strategy in strategies] |
|
792
|
|
|
return L[argmin([x.f for x in L])] |
|
793
|
|
|
|
|
794
|
|
|
class MultiStrategyDifferentialEvolution(DifferentialEvolution): |
|
795
|
|
|
r"""Implementation of Differential evolution algorithm with multiple mutation strateys. |
|
796
|
|
|
|
|
797
|
|
|
Algorithm: |
|
798
|
|
|
Implementation of Differential evolution algorithm with multiple mutation strateys |
|
799
|
|
|
|
|
800
|
|
|
Date: |
|
801
|
|
|
2018 |
|
802
|
|
|
|
|
803
|
|
|
Author: |
|
804
|
|
|
Klemen Berkovič |
|
805
|
|
|
|
|
806
|
|
|
License: |
|
807
|
|
|
MIT |
|
808
|
|
|
|
|
809
|
|
|
Attributes: |
|
810
|
|
|
Name (List[str]): List of strings representing algorithm names. |
|
811
|
|
|
strategies (Iterable[Callable[[numpy.ndarray[Individual], int, Individual, float, float, mtrand.RandomState], numpy.ndarray[Individual]]]): List of mutation strategies. |
|
812
|
|
|
CrossMutt (Callable[[numpy.ndarray[Individual], int, Individual, float, float, Task, Individual, Iterable[Callable[[numpy.ndarray, int, numpy.ndarray, float, float, mtrand.RandomState, Dict[str, Any]], Individual]]], Individual]): Multi crossover and mutation combiner function. |
|
813
|
|
|
|
|
814
|
|
|
See Also: |
|
815
|
|
|
* :class:`NiaPy.algorithms.basic.DifferentialEvolution` |
|
816
|
|
|
""" |
|
817
|
|
|
Name = ['MultiStrategyDifferentialEvolution', 'MsDE'] |
|
818
|
|
|
|
|
819
|
|
|
@staticmethod |
|
820
|
|
|
def algorithmInfo(): |
|
821
|
|
|
r"""Get basic information of algorithm. |
|
822
|
|
|
|
|
823
|
|
|
Returns: |
|
824
|
|
|
str: Basic information of algorithm. |
|
825
|
|
|
|
|
826
|
|
|
See Also: |
|
827
|
|
|
* :func:`NiaPy.algorithms.Algorithm.algorithmInfo` |
|
828
|
|
|
""" |
|
829
|
|
|
return r"""No info""" |
|
830
|
|
|
|
|
831
|
|
|
@staticmethod |
|
832
|
|
|
def typeParameters(): |
|
833
|
|
|
r"""Get dictionary with functions for checking values of parameters. |
|
834
|
|
|
|
|
835
|
|
|
Returns: |
|
836
|
|
|
Dict[str, Callable]: |
|
837
|
|
|
* CrossMutt (Callable[[Callable, bool]) |
|
838
|
|
|
|
|
839
|
|
|
See Also: |
|
840
|
|
|
* :func:`NiaPy.algorithms.basic.DifferentialEvolution.typeParameters` |
|
841
|
|
|
""" |
|
842
|
|
|
r = DifferentialEvolution.typeParameters() |
|
843
|
|
|
r.pop('CrossMutt', None) |
|
844
|
|
|
# TODO add constraint method for selection of stratgy methos |
|
845
|
|
|
return r |
|
846
|
|
|
|
|
847
|
|
|
def setParameters(self, strategies=(CrossRand1, CrossBest1, CrossCurr2Best1, CrossRand2), **ukwargs): |
|
848
|
|
|
r"""Set the arguments of the algorithm. |
|
849
|
|
|
|
|
850
|
|
|
Arguments: |
|
851
|
|
|
strategies (Optional[Iterable[Callable[[numpy.ndarray[Individual], int, Individual, float, float, mtrand.RandomState], numpy.ndarray[Individual]]]]): List of mutation strategyis. |
|
852
|
|
|
CrossMutt (Optional[Callable[[numpy.ndarray[Individual], int, Individual, float, float, Task, Individual, Iterable[Callable[[numpy.ndarray, int, numpy.ndarray, float, float, mtrand.RandomState, Dict[str, Any]], Individual]]], Individual]]): Multi crossover and mutation combiner function. |
|
853
|
|
|
|
|
854
|
|
|
See Also: |
|
855
|
|
|
* :func:`NiaPy.algorithms.basic.DifferentialEvolution.setParameters` |
|
856
|
|
|
""" |
|
857
|
|
|
DifferentialEvolution.setParameters(self, CrossMutt=multiMutations, **ukwargs) |
|
858
|
|
|
self.strategies = strategies |
|
859
|
|
|
|
|
860
|
|
View Code Duplication |
def evolve(self, pop, xb, task, **kwargs): |
|
|
|
|
|
|
861
|
|
|
r"""Evolve population with the help multiple mutation strategies. |
|
862
|
|
|
|
|
863
|
|
|
Args: |
|
864
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
865
|
|
|
xb (Individual): Current best individual. |
|
866
|
|
|
task (Task): Optimization task. |
|
867
|
|
|
**kwargs (Dict[str, Any]): Additional arguments. |
|
868
|
|
|
|
|
869
|
|
|
Returns: |
|
870
|
|
|
numpy.ndarray[Individual]: New population of individuals. |
|
871
|
|
|
""" |
|
872
|
|
|
return objects2array([self.CrossMutt(pop, i, xb, self.F, self.CR, self.Rand, task, self.itype, self.strategies) for i in range(len(pop))]) |
|
873
|
|
|
|
|
874
|
|
|
class DynNpMultiStrategyDifferentialEvolution(MultiStrategyDifferentialEvolution, DynNpDifferentialEvolution): |
|
875
|
|
|
r"""Implementation of Dynamic population size Differential evolution algorithm with dynamic population size that is defined by the quality of population. |
|
876
|
|
|
|
|
877
|
|
|
Algorithm: |
|
878
|
|
|
Dynamic population size Differential evolution algorithm with dynamic population size that is defined by the quality of population |
|
879
|
|
|
|
|
880
|
|
|
Date: |
|
881
|
|
|
2018 |
|
882
|
|
|
|
|
883
|
|
|
Author: |
|
884
|
|
|
Klemen Berkovič |
|
885
|
|
|
|
|
886
|
|
|
License: |
|
887
|
|
|
MIT |
|
888
|
|
|
|
|
889
|
|
|
Attributes: |
|
890
|
|
|
Name (List[str]): List of strings representing algorithm name. |
|
891
|
|
|
|
|
892
|
|
|
See Also: |
|
893
|
|
|
* :class:`NiaPy.algorithms.basic.MultiStrategyDifferentialEvolution` |
|
894
|
|
|
* :class:`NiaPy.algorithms.basic.DynNpDifferentialEvolution` |
|
895
|
|
|
""" |
|
896
|
|
|
Name = ['DynNpMultiStrategyDifferentialEvolution', 'dynNpMsDE'] |
|
897
|
|
|
|
|
898
|
|
|
@staticmethod |
|
899
|
|
|
def algorithmInfo(): |
|
900
|
|
|
r"""Get basic information of algorithm. |
|
901
|
|
|
|
|
902
|
|
|
Returns: |
|
903
|
|
|
str: Basic information of algorithm. |
|
904
|
|
|
|
|
905
|
|
|
See Also: |
|
906
|
|
|
* :func:`NiaPy.algorithms.Algorithm.algorithmInfo` |
|
907
|
|
|
""" |
|
908
|
|
|
return r"""No info""" |
|
909
|
|
|
|
|
910
|
|
|
@staticmethod |
|
911
|
|
|
def typeParameters(): |
|
912
|
|
|
r"""Get dictionary with functions for checking values of parameters. |
|
913
|
|
|
|
|
914
|
|
|
Returns: |
|
915
|
|
|
Dict[str, Callable]: |
|
916
|
|
|
* rp (Callable[[Union[float, int]], bool]): TODO |
|
917
|
|
|
* pmax (Callable[[int], bool]): TODO |
|
918
|
|
|
|
|
919
|
|
|
See Also: |
|
920
|
|
|
* :func:`NiaPy.algorithms.basic.MultiStrategyDifferentialEvolution.typeParameters` |
|
921
|
|
|
""" |
|
922
|
|
|
r = MultiStrategyDifferentialEvolution.typeParameters() |
|
923
|
|
|
r['rp'] = lambda x: isinstance(x, (float, int)) and x > 0 |
|
924
|
|
|
r['pmax'] = lambda x: isinstance(x, int) and x > 0 |
|
925
|
|
|
return r |
|
926
|
|
|
|
|
927
|
|
|
def setParameters(self, **ukwargs): |
|
928
|
|
|
r"""Set the arguments of the algorithm. |
|
929
|
|
|
|
|
930
|
|
|
Args: |
|
931
|
|
|
ukwargs (Dict[str, Any]): Additional arguments. |
|
932
|
|
|
|
|
933
|
|
|
See Also: |
|
934
|
|
|
* :func:`NiaPy.algorithms.basic.MultiStrategyDifferentialEvolution.setParameters` |
|
935
|
|
|
* :func:`NiaPy.algorithms.basic.DynNpDifferentialEvolution.setParameters` |
|
936
|
|
|
""" |
|
937
|
|
|
DynNpDifferentialEvolution.setParameters(self, **ukwargs) |
|
938
|
|
|
MultiStrategyDifferentialEvolution.setParameters(self, **ukwargs) |
|
939
|
|
|
|
|
940
|
|
|
def evolve(self, pop, xb, task, **kwargs): |
|
941
|
|
|
return MultiStrategyDifferentialEvolution.evolve(self, pop, xb, task, **kwargs) |
|
942
|
|
|
|
|
943
|
|
|
def postSelection(self, pop, task, **kwargs): |
|
944
|
|
|
r"""Post selection operator. |
|
945
|
|
|
|
|
946
|
|
|
Args: |
|
947
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
948
|
|
|
task (Task): Optimization task. |
|
949
|
|
|
**kwargs (Dict[str, Any]): Additional arguments. |
|
950
|
|
|
|
|
951
|
|
|
Returns: |
|
952
|
|
|
numpy.ndarray: New population. |
|
953
|
|
|
|
|
954
|
|
|
See Also: |
|
955
|
|
|
* :func:`NiaPy.algorithms.basic.DynNpDifferentialEvolution.postSelection` |
|
956
|
|
|
""" |
|
957
|
|
|
return DynNpDifferentialEvolution.postSelection(self, pop, task) |
|
958
|
|
|
|
|
959
|
|
|
class AgingNpMultiMutationDifferentialEvolution(AgingNpDifferentialEvolution, MultiStrategyDifferentialEvolution): |
|
960
|
|
|
r"""Implementation of Differential evolution algorithm with aging individuals. |
|
961
|
|
|
|
|
962
|
|
|
Algorithm: |
|
963
|
|
|
Differential evolution algorithm with dynamic population size that is defined by the quality of population |
|
964
|
|
|
|
|
965
|
|
|
Date: |
|
966
|
|
|
2018 |
|
967
|
|
|
|
|
968
|
|
|
Author: |
|
969
|
|
|
Klemen Berkovič |
|
970
|
|
|
|
|
971
|
|
|
License: |
|
972
|
|
|
MIT |
|
973
|
|
|
|
|
974
|
|
|
Attributes: |
|
975
|
|
|
Name (List[str]): List of strings representing algorithm names |
|
976
|
|
|
|
|
977
|
|
|
See Also: |
|
978
|
|
|
* :class:`NiaPy.algorithms.basic.AgingNpDifferentialEvolution` |
|
979
|
|
|
* :class:`NiaPy.algorithms.basic.MultiStrategyDifferentialEvolution` |
|
980
|
|
|
""" |
|
981
|
|
|
Name = ['AgingNpMultiMutationDifferentialEvolution', 'ANpMSDE'] |
|
982
|
|
|
|
|
983
|
|
|
@staticmethod |
|
984
|
|
|
def algorithmInfo(): |
|
985
|
|
|
r"""Get basic information of algorithm. |
|
986
|
|
|
|
|
987
|
|
|
Returns: |
|
988
|
|
|
str: Basic information of algorithm. |
|
989
|
|
|
|
|
990
|
|
|
See Also: |
|
991
|
|
|
* :func:`NiaPy.algorithms.Algorithm.algorithmInfo` |
|
992
|
|
|
""" |
|
993
|
|
|
return r"""No info""" |
|
994
|
|
|
|
|
995
|
|
|
@staticmethod |
|
996
|
|
|
def typeParameters(): |
|
997
|
|
|
r"""Get dictionary with functions for checking values of parameters. |
|
998
|
|
|
|
|
999
|
|
|
Returns: |
|
1000
|
|
|
Dict[str, Callable]: |
|
1001
|
|
|
* rp (Callable[[Union[float, int]], bool]): TODO |
|
1002
|
|
|
* pmax (Callable[[int], bool]): TODO |
|
1003
|
|
|
|
|
1004
|
|
|
See Also: |
|
1005
|
|
|
* :func:`NiaPy.algorithms.basic.MultiStrategyDifferentialEvolution.typeParameters` |
|
1006
|
|
|
""" |
|
1007
|
|
|
r = AgingNpDifferentialEvolution.typeParameters() |
|
1008
|
|
|
# TODO add other parameters to data check list |
|
1009
|
|
|
return r |
|
1010
|
|
|
|
|
1011
|
|
|
def setParameters(self, **ukwargs): |
|
1012
|
|
|
r"""Set core parameter arguments. |
|
1013
|
|
|
|
|
1014
|
|
|
Args: |
|
1015
|
|
|
**ukwargs (Dict[str, Any]): Additional arguments. |
|
1016
|
|
|
|
|
1017
|
|
|
See Also: |
|
1018
|
|
|
* :func:`NiaPy.algorithms.basic.AgingNpDifferentialEvolution.setParameters` |
|
1019
|
|
|
* :func:`NiaPy.algorithms.basic.MultiStrategyDifferentialEvolution.setParameters` |
|
1020
|
|
|
""" |
|
1021
|
|
|
AgingNpDifferentialEvolution.setParameters(self, **ukwargs) |
|
1022
|
|
|
MultiStrategyDifferentialEvolution.setParameters(self, stratgeys=(CrossRand1, CrossBest1, CrossCurr2Rand1, CrossRand2), itype=AgingIndividual, **ukwargs) |
|
1023
|
|
|
|
|
1024
|
|
|
def evolve(self, pop, xb, task, **kwargs): |
|
1025
|
|
|
r"""Evolve current population. |
|
1026
|
|
|
|
|
1027
|
|
|
Args: |
|
1028
|
|
|
pop (numpy.ndarray[Individual]): Current population. |
|
1029
|
|
|
xb (Individual): Global best individual. |
|
1030
|
|
|
task (Task): Optimization task. |
|
1031
|
|
|
**kwargs (Dict[str, Any]): Additional arguments. |
|
1032
|
|
|
|
|
1033
|
|
|
Returns: |
|
1034
|
|
|
numpy.ndarray[Individual]: New population of individuals. |
|
1035
|
|
|
""" |
|
1036
|
|
|
return MultiStrategyDifferentialEvolution.evolve(self, pop, xb, task, **kwargs) |
|
1037
|
|
|
|
|
1038
|
|
|
# vim: tabstop=3 noexpandtab shiftwidth=3 softtabstop=3 |
|
1039
|
|
|
|
NiaOrg /
NiaPy
Loading history...