|
1
|
|
|
# encoding=utf8 |
|
2
|
|
|
# pylint: disable=mixed-indentation, trailing-whitespace, multiple-statements, attribute-defined-outside-init, logging-not-lazy, no-self-use, len-as-condition, singleton-comparison, arguments-differ, line-too-long, unused-argument, consider-using-enumerate, bad-continuation, superfluous-parens, redefined-builtin |
|
3
|
|
|
import logging |
|
4
|
|
|
import operator as oper |
|
5
|
|
|
|
|
6
|
|
|
from numpy import random as rand, vectorize, where, copy, apply_along_axis, argmin, argmax, argsort, fmin, fmax, full, asarray, abs, inf |
|
7
|
|
|
|
|
8
|
|
|
from NiaPy.algorithms.algorithm import Algorithm |
|
9
|
|
|
|
|
10
|
|
|
logging.basicConfig() |
|
11
|
|
|
logger = logging.getLogger('NiaPy.algorithms.other') |
|
12
|
|
|
logger.setLevel('INFO') |
|
13
|
|
|
|
|
14
|
|
|
__all__ = ['MultipleTrajectorySearch', 'MultipleTrajectorySearchV1', 'MTS_LS1', 'MTS_LS1v1', 'MTS_LS2', 'MTS_LS3', 'MTS_LS3v1'] |
|
15
|
|
|
|
|
16
|
|
View Code Duplication |
def MTS_LS1(Xk, Xk_fit, Xb, Xb_fit, improve, SR, task, BONUS1=10, BONUS2=1, sr_fix=0.4, rnd=rand, **ukwargs): |
|
|
|
|
|
|
17
|
|
|
r"""Multiple trajectory local search one. |
|
18
|
|
|
|
|
19
|
|
|
Args: |
|
20
|
|
|
Xk (numpy.ndarray): Current solution. |
|
21
|
|
|
Xk_fit (float): Current solutions fitness/function value. |
|
22
|
|
|
Xb (numpy.ndarray): Global best solution. |
|
23
|
|
|
Xb_fit (float): Global best solutions fitness/function value. |
|
24
|
|
|
improve (bool): Has the solution been improved. |
|
25
|
|
|
SR (numpy.ndarray): Search range. |
|
26
|
|
|
task (Task): Optimization task. |
|
27
|
|
|
BONUS1 (int): Bonus reward for improving global best solution. |
|
28
|
|
|
BONUS2 (int): Bonus reward for improving solution. |
|
29
|
|
|
sr_fix (numpy.ndarray): Fix when search range is to small. |
|
30
|
|
|
rnd (mtrand.RandomState): Random number generator. |
|
31
|
|
|
**ukwargs (Dict[str, Any]): Additional arguments. |
|
32
|
|
|
|
|
33
|
|
|
Returns: |
|
34
|
|
|
Tuple[numpy.ndarray, float, numpy.ndarray, float, bool, numpy.ndarray]: |
|
35
|
|
|
1. New solution. |
|
36
|
|
|
2. New solutions fitness/function value. |
|
37
|
|
|
3. Global best if found else old global best. |
|
38
|
|
|
4. Global bests function/fitness value. |
|
39
|
|
|
5. If solution has improved. |
|
40
|
|
|
6. Search range. |
|
41
|
|
|
""" |
|
42
|
|
|
if not improve: |
|
43
|
|
|
SR /= 2 |
|
44
|
|
|
ifix = where(SR < 1e-15) |
|
45
|
|
|
SR[ifix] = task.bRange[ifix] * sr_fix |
|
46
|
|
|
improve, grade = False, 0.0 |
|
47
|
|
|
for i in range(len(Xk)): |
|
48
|
|
|
Xk_i_old = Xk[i] |
|
49
|
|
|
Xk[i] = Xk_i_old - SR[i] |
|
50
|
|
|
Xk = task.repair(Xk, rnd) |
|
51
|
|
|
Xk_fit_new = task.eval(Xk) |
|
52
|
|
|
if Xk_fit_new < Xb_fit: grade, Xb, Xb_fit = grade + BONUS1, Xk, Xk_fit_new |
|
53
|
|
|
if Xk_fit_new == Xk_fit: Xk[i] = Xk_i_old |
|
54
|
|
|
elif Xk_fit_new > Xk_fit: |
|
55
|
|
|
Xk[i] = Xk_i_old + 0.5 * SR[i] |
|
56
|
|
|
Xk = task.repair(Xk, rnd) |
|
57
|
|
|
Xk_fit_new = task.eval(task.repair(Xk, rnd)) |
|
58
|
|
|
if Xk_fit_new < Xb_fit: grade, Xb, Xb_fit = grade + BONUS1, copy(Xk), Xk_fit_new |
|
59
|
|
|
if Xk_fit_new >= Xk_fit: Xk[i] = Xk_i_old |
|
60
|
|
|
else: grade, improve, Xk_fit = grade + BONUS2, True, Xk_fit_new |
|
61
|
|
|
else: grade, improve, Xk_fit = grade + BONUS2, True, Xk_fit_new |
|
62
|
|
|
return Xk, Xk_fit, Xb, Xb_fit, improve, grade, SR |
|
63
|
|
|
|
|
64
|
|
View Code Duplication |
def MTS_LS1v1(Xk, Xk_fit, Xb, Xb_fit, improve, SR, task, BONUS1=10, BONUS2=1, sr_fix=0.4, rnd=rand, **ukwargs): |
|
|
|
|
|
|
65
|
|
|
r"""Multiple trajectory local search one version two. |
|
66
|
|
|
|
|
67
|
|
|
Args: |
|
68
|
|
|
Xk (numpy.ndarray): Current solution. |
|
69
|
|
|
Xk_fit (float): Current solutions fitness/function value. |
|
70
|
|
|
Xb (numpy.ndarray): Global best solution. |
|
71
|
|
|
Xb_fit (float): Global best solutions fitness/function value. |
|
72
|
|
|
improve (bool): Has the solution been improved. |
|
73
|
|
|
SR (numpy.ndarray): Search range. |
|
74
|
|
|
task (Task): Optimization task. |
|
75
|
|
|
BONUS1 (int): Bonus reward for improving global best solution. |
|
76
|
|
|
BONUS2 (int): Bonus reward for improving solution. |
|
77
|
|
|
sr_fix (numpy.ndarray): Fix when search range is to small. |
|
78
|
|
|
rnd (mtrand.RandomState): Random number generator. |
|
79
|
|
|
**ukwargs (Dict[str, Any]): Additional arguments. |
|
80
|
|
|
|
|
81
|
|
|
Returns: |
|
82
|
|
|
Tuple[numpy.ndarray, float, numpy.ndarray, float, bool, numpy.ndarray]: |
|
83
|
|
|
1. New solution. |
|
84
|
|
|
2. New solutions fitness/function value. |
|
85
|
|
|
3. Global best if found else old global best. |
|
86
|
|
|
4. Global bests function/fitness value. |
|
87
|
|
|
5. If solution has improved. |
|
88
|
|
|
6. Search range. |
|
89
|
|
|
""" |
|
90
|
|
|
if not improve: |
|
91
|
|
|
SR /= 2 |
|
92
|
|
|
ifix = where(SR < 1e-15) |
|
93
|
|
|
SR[ifix] = task.bRange[ifix] * sr_fix |
|
94
|
|
|
improve, D, grade = False, rnd.uniform(-1, 1, task.D), 0.0 |
|
95
|
|
|
for i in range(len(Xk)): |
|
96
|
|
|
Xk_i_old = Xk[i] |
|
97
|
|
|
Xk[i] = Xk_i_old - SR[i] * D[i] |
|
98
|
|
|
Xk = task.repair(Xk, rnd) |
|
99
|
|
|
Xk_fit_new = task.eval(Xk) |
|
100
|
|
|
if Xk_fit_new < Xb_fit: grade, Xb, Xb_fit = grade + BONUS1, Xk, Xk_fit_new |
|
101
|
|
|
elif Xk_fit_new == Xk_fit: Xk[i] = Xk_i_old |
|
102
|
|
|
elif Xk_fit_new > Xk_fit: |
|
103
|
|
|
Xk[i] = Xk_i_old + 0.5 * SR[i] |
|
104
|
|
|
Xk = task.repair(Xk, rnd) |
|
105
|
|
|
Xk_fit_new = task.eval(Xk) |
|
106
|
|
|
if Xk_fit_new < Xb_fit: grade, Xb, Xb_fit = grade + BONUS1, Xk, Xk_fit_new |
|
107
|
|
|
if Xk_fit_new >= Xk_fit: Xk[i] = Xk_i_old |
|
108
|
|
|
else: grade, improve, Xk_fit = grade + BONUS2, True, Xk_fit_new |
|
109
|
|
|
else: grade, improve, Xk_fit = grade + BONUS2, True, Xk_fit_new |
|
110
|
|
|
return Xk, Xk_fit, Xb, Xb_fit, improve, grade, SR |
|
111
|
|
|
|
|
112
|
|
|
def genNewX(x, r, d, SR, op): |
|
113
|
|
|
r"""Move solution to other position based on operator. |
|
114
|
|
|
|
|
115
|
|
|
Args: |
|
116
|
|
|
x (numpy.ndarray): Solution to move. |
|
117
|
|
|
r (int): Random number. |
|
118
|
|
|
d (float): Scale factor. |
|
119
|
|
|
SR (numpy.ndarray): Search range. |
|
120
|
|
|
op (operator): Operator to use. |
|
121
|
|
|
|
|
122
|
|
|
Returns: |
|
123
|
|
|
numpy.ndarray: Moved solution based on operator. |
|
124
|
|
|
""" |
|
125
|
|
|
return op(x, SR * d) if r == 0 else x |
|
126
|
|
|
|
|
127
|
|
|
def MTS_LS2(Xk, Xk_fit, Xb, Xb_fit, improve, SR, task, BONUS1=10, BONUS2=1, sr_fix=0.4, rnd=rand, **ukwargs): |
|
128
|
|
|
r"""Multiple trajectory local search two. |
|
129
|
|
|
|
|
130
|
|
|
Args: |
|
131
|
|
|
Xk (numpy.ndarray): Current solution. |
|
132
|
|
|
Xk_fit (float): Current solutions fitness/function value. |
|
133
|
|
|
Xb (numpy.ndarray): Global best solution. |
|
134
|
|
|
Xb_fit (float): Global best solutions fitness/function value. |
|
135
|
|
|
improve (bool): Has the solution been improved. |
|
136
|
|
|
SR (numpy.ndarray): Search range. |
|
137
|
|
|
task (Task): Optimization task. |
|
138
|
|
|
BONUS1 (int): Bonus reward for improving global best solution. |
|
139
|
|
|
BONUS2 (int): Bonus reward for improving solution. |
|
140
|
|
|
sr_fix (numpy.ndarray): Fix when search range is to small. |
|
141
|
|
|
rnd (mtrand.RandomState): Random number generator. |
|
142
|
|
|
**ukwargs (Dict[str, Any]): Additional arguments. |
|
143
|
|
|
|
|
144
|
|
|
Returns: |
|
145
|
|
|
Tuple[numpy.ndarray, float, numpy.ndarray, float, bool, numpy.ndarray]: |
|
146
|
|
|
1. New solution. |
|
147
|
|
|
2. New solutions fitness/function value. |
|
148
|
|
|
3. Global best if found else old global best. |
|
149
|
|
|
4. Global bests function/fitness value. |
|
150
|
|
|
5. If solution has improved. |
|
151
|
|
|
6. Search range. |
|
152
|
|
|
|
|
153
|
|
|
See Also: |
|
154
|
|
|
* :func:`NiaPy.algorithms.other.genNewX` |
|
155
|
|
|
""" |
|
156
|
|
|
if not improve: |
|
157
|
|
|
SR /= 2 |
|
158
|
|
|
ifix = where(SR < 1e-15) |
|
159
|
|
|
SR[ifix] = task.bRange[ifix] * sr_fix |
|
160
|
|
|
improve, grade = False, 0.0 |
|
161
|
|
|
for _ in range(len(Xk)): |
|
162
|
|
|
D = -1 + rnd.rand(len(Xk)) * 2 |
|
163
|
|
|
R = rnd.choice([0, 1, 2, 3], len(Xk)) |
|
164
|
|
|
Xk_new = task.repair(vectorize(genNewX)(Xk, R, D, SR, oper.sub), rnd) |
|
165
|
|
|
Xk_fit_new = task.eval(Xk_new) |
|
166
|
|
|
if Xk_fit_new < Xb_fit: grade, Xb, Xb_fit = grade + BONUS1, Xk_new, Xk_fit_new |
|
167
|
|
|
elif Xk_fit_new != Xk_fit: |
|
168
|
|
|
if Xk_fit_new > Xk_fit: |
|
169
|
|
|
Xk_new = task.repair(vectorize(genNewX)(Xk, R, D, SR, oper.add), rnd) |
|
170
|
|
|
Xk_fit_new = task.eval(Xk_new) |
|
171
|
|
|
if Xk_fit_new < Xb_fit: grade, Xb, Xb_fit = grade + BONUS1, Xk_new, Xk_fit_new |
|
172
|
|
|
if Xk_fit_new < Xk_fit: grade, Xk, Xk_fit, improve = grade + BONUS2, Xk_new, Xk_fit_new, True |
|
173
|
|
|
else: grade, Xk, Xk_fit, improve = grade + BONUS2, Xk_new, Xk_fit_new, True |
|
174
|
|
|
return Xk, Xk_fit, Xb, Xb_fit, improve, grade, SR |
|
175
|
|
|
|
|
176
|
|
|
def MTS_LS3(Xk, Xk_fit, Xb, Xb_fit, improve, SR, task, BONUS1=10, BONUS2=1, rnd=rand, **ukwargs): |
|
177
|
|
|
r"""Multiple trajectory local search three. |
|
178
|
|
|
|
|
179
|
|
|
Args: |
|
180
|
|
|
Xk (numpy.ndarray): Current solution. |
|
181
|
|
|
Xk_fit (float): Current solutions fitness/function value. |
|
182
|
|
|
Xb (numpy.ndarray): Global best solution. |
|
183
|
|
|
Xb_fit (float): Global best solutions fitness/function value. |
|
184
|
|
|
improve (bool): Has the solution been improved. |
|
185
|
|
|
SR (numpy.ndarray): Search range. |
|
186
|
|
|
task (Task): Optimization task. |
|
187
|
|
|
BONUS1 (int): Bonus reward for improving global best solution. |
|
188
|
|
|
BONUS2 (int): Bonus reward for improving solution. |
|
189
|
|
|
rnd (mtrand.RandomState): Random number generator. |
|
190
|
|
|
**ukwargs (Dict[str, Any]): Additional arguments. |
|
191
|
|
|
|
|
192
|
|
|
Returns: |
|
193
|
|
|
Tuple[numpy.ndarray, float, numpy.ndarray, float, bool, numpy.ndarray]: |
|
194
|
|
|
1. New solution. |
|
195
|
|
|
2. New solutions fitness/function value. |
|
196
|
|
|
3. Global best if found else old global best. |
|
197
|
|
|
4. Global bests function/fitness value. |
|
198
|
|
|
5. If solution has improved. |
|
199
|
|
|
6. Search range. |
|
200
|
|
|
""" |
|
201
|
|
|
Xk_new, grade = copy(Xk), 0.0 |
|
202
|
|
|
for i in range(len(Xk)): |
|
203
|
|
|
Xk1, Xk2, Xk3 = copy(Xk_new), copy(Xk_new), copy(Xk_new) |
|
204
|
|
|
Xk1[i], Xk2[i], Xk3[i] = Xk1[i] + 0.1, Xk2[i] - 0.1, Xk3[i] + 0.2 |
|
205
|
|
|
Xk1, Xk2, Xk3 = task.repair(Xk1, rnd), task.repair(Xk2, rnd), task.repair(Xk3, rnd) |
|
206
|
|
|
Xk1_fit, Xk2_fit, Xk3_fit = task.eval(Xk1), task.eval(Xk2), task.eval(Xk3) |
|
207
|
|
|
if Xk1_fit < Xb_fit: grade, Xb, Xb_fit, improve = grade + BONUS1, Xk1, Xk1_fit, True |
|
208
|
|
|
if Xk2_fit < Xb_fit: grade, Xb, Xb_fit, improve = grade + BONUS1, Xk2, Xk2_fit, True |
|
209
|
|
|
if Xk3_fit < Xb_fit: grade, Xb, Xb_fit, improve = grade + BONUS1, Xk3, Xk3_fit, True |
|
210
|
|
|
D1, D2, D3 = Xk_fit - Xk1_fit if abs(Xk1_fit) != inf else 0, Xk_fit - Xk2_fit if abs(Xk2_fit) != inf else 0, Xk_fit - Xk3_fit if abs(Xk3_fit) != inf else 0 |
|
211
|
|
|
if D1 > 0: grade, improve = grade + BONUS2, True |
|
212
|
|
|
if D2 > 0: grade, improve = grade + BONUS2, True |
|
213
|
|
|
if D3 > 0: grade, improve = grade + BONUS2, True |
|
214
|
|
|
a, b, c = 0.4 + rnd.rand() * 0.1, 0.1 + rnd.rand() * 0.2, rnd.rand() |
|
215
|
|
|
Xk_new[i] += a * (D1 - D2) + b * (D3 - 2 * D1) + c |
|
216
|
|
|
Xk_new = task.repair(Xk_new, rnd) |
|
217
|
|
|
Xk_fit_new = task.eval(Xk_new) |
|
218
|
|
|
if Xk_fit_new < Xk_fit: Xk, Xk_fit, improve = Xk_new, Xk_fit_new, True |
|
219
|
|
|
return Xk, Xk_fit, Xb, Xb_fit, improve, grade, SR |
|
220
|
|
|
|
|
221
|
|
|
def MTS_LS3v1(Xk, Xk_fit, Xb, Xb_fit, improve, SR, task, phi=3, BONUS1=10, BONUS2=1, rnd=rand, **ukwargs): |
|
222
|
|
|
r"""Multiple trajectory local search three version one. |
|
223
|
|
|
|
|
224
|
|
|
Args: |
|
225
|
|
|
Xk (numpy.ndarray): Current solution. |
|
226
|
|
|
Xk_fit (float): Current solutions fitness/function value. |
|
227
|
|
|
Xb (numpy.ndarray): Global best solution. |
|
228
|
|
|
Xb_fit (float): Global best solutions fitness/function value. |
|
229
|
|
|
improve (bool): Has the solution been improved. |
|
230
|
|
|
SR (numpy.ndarray): Search range. |
|
231
|
|
|
task (Task): Optimization task. |
|
232
|
|
|
phi (int): Number of new generated positions. |
|
233
|
|
|
BONUS1 (int): Bonus reward for improving global best solution. |
|
234
|
|
|
BONUS2 (int): Bonus reward for improving solution. |
|
235
|
|
|
rnd (mtrand.RandomState): Random number generator. |
|
236
|
|
|
**ukwargs (Dict[str, Any]): Additional arguments. |
|
237
|
|
|
|
|
238
|
|
|
Returns: |
|
239
|
|
|
Tuple[numpy.ndarray, float, numpy.ndarray, float, bool, numpy.ndarray]: |
|
240
|
|
|
1. New solution. |
|
241
|
|
|
2. New solutions fitness/function value. |
|
242
|
|
|
3. Global best if found else old global best. |
|
243
|
|
|
4. Global bests function/fitness value. |
|
244
|
|
|
5. If solution has improved. |
|
245
|
|
|
6. Search range. |
|
246
|
|
|
""" |
|
247
|
|
|
grade, Disp = 0.0, task.bRange / 10 |
|
248
|
|
|
while True in (Disp > 1e-3): |
|
249
|
|
|
Xn = apply_along_axis(task.repair, 1, asarray([rnd.permutation(Xk) + Disp * rnd.uniform(-1, 1, len(Xk)) for _ in range(phi)]), rnd) |
|
250
|
|
|
Xn_f = apply_along_axis(task.eval, 1, Xn) |
|
251
|
|
|
iBetter, iBetterBest = where(Xn_f < Xk_fit), where(Xn_f < Xb_fit) |
|
252
|
|
|
grade += len(iBetterBest) * BONUS1 + (len(iBetter) - len(iBetterBest)) * BONUS2 |
|
253
|
|
|
if len(Xn_f[iBetterBest]) > 0: |
|
254
|
|
|
ib, improve = argmin(Xn_f[iBetterBest]), True |
|
255
|
|
|
Xb, Xb_fit, Xk, Xk_fit = Xn[ib], Xn_f[ib], Xn[ib], Xn_f[ib] |
|
256
|
|
|
elif len(Xn_f[iBetter]) > 0: |
|
257
|
|
|
ib, improve = argmin(Xn_f[iBetter]), True |
|
258
|
|
|
Xk, Xk_fit = Xn[ib], Xn_f[ib] |
|
259
|
|
|
Su, Sl = fmin(task.Upper, Xk + 2 * Disp), fmax(task.Lower, Xk - 2 * Disp) |
|
260
|
|
|
Disp = (Su - Sl) / 10 |
|
261
|
|
|
return Xk, Xk_fit, Xb, Xb_fit, improve, grade, SR |
|
262
|
|
|
|
|
263
|
|
|
class MultipleTrajectorySearch(Algorithm): |
|
264
|
|
|
r"""Implementation of Multiple trajectory search. |
|
265
|
|
|
|
|
266
|
|
|
Algorithm: |
|
267
|
|
|
Multiple trajectory search |
|
268
|
|
|
|
|
269
|
|
|
Date: |
|
270
|
|
|
2018 |
|
271
|
|
|
|
|
272
|
|
|
Authors: |
|
273
|
|
|
Klemen Berkovic |
|
274
|
|
|
|
|
275
|
|
|
License: |
|
276
|
|
|
MIT |
|
277
|
|
|
|
|
278
|
|
|
Reference URL: |
|
279
|
|
|
https://ieeexplore.ieee.org/document/4631210/ |
|
280
|
|
|
|
|
281
|
|
|
Reference paper: |
|
282
|
|
|
Lin-Yu Tseng and Chun Chen, "Multiple trajectory search for Large Scale Global Optimization," 2008 IEEE Congress on Evolutionary Computation (IEEE World Congress on Computational Intelligence), Hong Kong, 2008, pp. 3052-3059. doi: 10.1109/CEC.2008.4631210 |
|
283
|
|
|
|
|
284
|
|
|
Attributes: |
|
285
|
|
|
Name (List[Str]): List of strings representing algorithm name. |
|
286
|
|
|
LSs (Iterable[Callable[[numpy.ndarray, float, numpy.ndarray, float, bool, numpy.ndarray, Task, Dict[str, Any]], Tuple[numpy.ndarray, float, numpy.ndarray, float, bool, int, numpy.ndarray]]]): Local searches to use. |
|
287
|
|
|
BONUS1 (int): Bonus for improving global best solution. |
|
288
|
|
|
BONUS2 (int): Bonus for improving solution. |
|
289
|
|
|
NoLsTests (int): Number of test runs on local search algorithms. |
|
290
|
|
|
NoLs (int): Number of local search algorithm runs. |
|
291
|
|
|
NoLsBest (int): Number of locals search algorithm runs on best solution. |
|
292
|
|
|
NoEnabled (int): Number of best solution for testing. |
|
293
|
|
|
|
|
294
|
|
|
See Also: |
|
295
|
|
|
* :class:`NiaPy.algorithms.Algorithm` |
|
296
|
|
|
""" |
|
297
|
|
|
Name = ['MultipleTrajectorySearch', 'MTS'] |
|
298
|
|
|
|
|
299
|
|
|
@staticmethod |
|
300
|
|
|
def algorithmInfo(): |
|
301
|
|
|
r"""Get basic information of algorithm. |
|
302
|
|
|
|
|
303
|
|
|
Returns: |
|
304
|
|
|
str: Basic information of algorithm. |
|
305
|
|
|
|
|
306
|
|
|
See Also: |
|
307
|
|
|
* :func:`NiaPy.algorithms.Algorithm.algorithmInfo` |
|
308
|
|
|
""" |
|
309
|
|
|
return r"""Lin-Yu Tseng and Chun Chen, "Multiple trajectory search for Large Scale Global Optimization," 2008 IEEE Congress on Evolutionary Computation (IEEE World Congress on Computational Intelligence), Hong Kong, 2008, pp. 3052-3059. doi: 10.1109/CEC.2008.4631210""" |
|
310
|
|
|
|
|
311
|
|
View Code Duplication |
@staticmethod |
|
|
|
|
|
|
312
|
|
|
def typeParameters(): |
|
313
|
|
|
r"""Get dictionary with functions for checking values of parameters. |
|
314
|
|
|
|
|
315
|
|
|
Returns: |
|
316
|
|
|
Dict[str, Callable]: |
|
317
|
|
|
* NoLsTests (Callable[[int], bool]) |
|
318
|
|
|
* NoLs (Callable[[int], bool]) |
|
319
|
|
|
* NoLsBest (Callable[[int], bool]) |
|
320
|
|
|
* NoEnabled (Callable[[int], bool]) |
|
321
|
|
|
""" |
|
322
|
|
|
return { |
|
323
|
|
|
'NoLsTests': lambda x: isinstance(x, int) and x >= 0, |
|
324
|
|
|
'NoLs': lambda x: isinstance(x, int) and x >= 0, |
|
325
|
|
|
'NoLsBest': lambda x: isinstance(x, int) and x >= 0, |
|
326
|
|
|
'NoEnabled': lambda x: isinstance(x, int) and x > 0 |
|
327
|
|
|
} |
|
328
|
|
|
|
|
329
|
|
|
def setParameters(self, M=40, NoLsTests=5, NoLs=5, NoLsBest=5, NoEnabled=17, BONUS1=10, BONUS2=1, LSs=(MTS_LS1v1, MTS_LS2, MTS_LS3v1), **ukwargs): |
|
330
|
|
|
r"""Set the arguments of the algorithm. |
|
331
|
|
|
|
|
332
|
|
|
Arguments: |
|
333
|
|
|
M (int): Number of individuals in population. |
|
334
|
|
|
NoLsTests (int): Number of test runs on local search algorithms. |
|
335
|
|
|
NoLs (int): Number of local search algorithm runs. |
|
336
|
|
|
NoLsBest (int): Number of locals search algorithm runs on best solution. |
|
337
|
|
|
NoEnabled (int): Number of best solution for testing. |
|
338
|
|
|
BONUS1 (int): Bonus for improving global best solution. |
|
339
|
|
|
BONUS2 (int): Bonus for improving self. |
|
340
|
|
|
LSs (Iterable[Callable[[numpy.ndarray, float, numpy.ndarray, float, bool, numpy.ndarray, Task, Dict[str, Any]], Tuple[numpy.ndarray, float, numpy.ndarray, float, bool, int, numpy.ndarray]]]): Local searches to use. |
|
341
|
|
|
|
|
342
|
|
|
See Also: |
|
343
|
|
|
* :func:`NiaPy.algorithms.Algorithm.setParameters` |
|
344
|
|
|
""" |
|
345
|
|
|
Algorithm.setParameters(self, NP=M) |
|
346
|
|
|
self.NoLsTests, self.NoLs, self.NoLsBest, self.NoEnabled, self.BONUS1, self.BONUS2 = NoLsTests, NoLs, NoLsBest, NoEnabled, BONUS1, BONUS2 |
|
347
|
|
|
self.LSs = LSs |
|
348
|
|
|
if ukwargs: logger.info('Unused arguments: %s' % (ukwargs)) |
|
349
|
|
|
|
|
350
|
|
|
def GradingRun(self, x, x_f, xb, xb_f, improve, SR, task): |
|
351
|
|
|
r"""Run local search for getting scores of local searches. |
|
352
|
|
|
|
|
353
|
|
|
Args: |
|
354
|
|
|
x (numpy.ndarray): Solution for grading. |
|
355
|
|
|
x_f (float): Solutions fitness/function value. |
|
356
|
|
|
xb (numpy.ndarray): Global best solution. |
|
357
|
|
|
xb_f (float): Global best solutions function/fitness value. |
|
358
|
|
|
improve (bool): Info if solution has improved. |
|
359
|
|
|
SR (numpy.ndarray): Search range. |
|
360
|
|
|
task (Task): Optimization task. |
|
361
|
|
|
|
|
362
|
|
|
Returns: |
|
363
|
|
|
Tuple[numpy.ndarray, float, numpy.ndarray, float]: |
|
364
|
|
|
1. New solution. |
|
365
|
|
|
2. New solutions function/fitness value. |
|
366
|
|
|
3. Global best solution. |
|
367
|
|
|
4. Global best solutions fitness/function value. |
|
368
|
|
|
""" |
|
369
|
|
|
ls_grades, Xn, Xnb = full(3, 0.0), [[x, x_f]] * len(self.LSs), [xb, xb_f] |
|
370
|
|
|
for _ in range(self.NoLsTests): |
|
371
|
|
|
for k in range(len(self.LSs)): |
|
372
|
|
|
Xn[k][0], Xn[k][1], xnb, xnb_f, improve, g, SR = self.LSs[k](Xn[k][0], Xn[k][1], Xnb[0], Xnb[1], improve, SR, task, BONUS1=self.BONUS1, BONUS2=self.BONUS2, rnd=self.Rand) |
|
373
|
|
|
if Xnb[1] > xnb_f: Xnb = [xnb, xnb_f] |
|
374
|
|
|
ls_grades[k] += g |
|
375
|
|
|
xn, k = min(Xn, key=lambda x: x[1]), argmax(ls_grades) |
|
376
|
|
|
return xn[0], xn[1], Xnb[0], Xnb[1], k |
|
377
|
|
|
|
|
378
|
|
|
def LsRun(self, k, x, x_f, xb, xb_f, improve, SR, g, task): |
|
379
|
|
|
r"""Run a selected local search. |
|
380
|
|
|
|
|
381
|
|
|
Args: |
|
382
|
|
|
k (int): Index of local search. |
|
383
|
|
|
x (numpy.ndarray): Current solution. |
|
384
|
|
|
x_f (float): Current solutions function/fitness value. |
|
385
|
|
|
xb (numpy.ndarray): Global best solution. |
|
386
|
|
|
xb_f (float): Global best solutions fitness/function value. |
|
387
|
|
|
improve (bool): If the solution has improved. |
|
388
|
|
|
SR (numpy.ndarray): Search range. |
|
389
|
|
|
g: |
|
390
|
|
|
task (Task): Optimization task. |
|
391
|
|
|
|
|
392
|
|
|
Returns: |
|
393
|
|
|
Tuple[numpy.ndarray, float, numpy.ndarray, float, bool, numpy.ndarray, int]: |
|
394
|
|
|
1. New best solution found. |
|
395
|
|
|
2. New best solutions found function/fitness value. |
|
396
|
|
|
3. Global best solution. |
|
397
|
|
|
4. Global best solutions function/fitness value. |
|
398
|
|
|
5. If the solution has improved. |
|
399
|
|
|
6. Grade of local search run. |
|
400
|
|
|
""" |
|
401
|
|
|
XBn = list() |
|
402
|
|
|
for _j in range(self.NoLs): |
|
403
|
|
|
x, x_f, xnb, xnb_f, improve, grade, SR = self.LSs[k](x, x_f, xb, xb_f, improve, SR, task, BONUS1=self.BONUS1, BONUS2=self.BONUS2, rnd=self.Rand) |
|
404
|
|
|
g += grade |
|
405
|
|
|
XBn.append((xnb, xnb_f)) |
|
406
|
|
|
xb, xb_f = min(XBn, key=lambda x: x[1]) |
|
407
|
|
|
return x, x_f, xb, xb_f, improve, SR, g |
|
408
|
|
|
|
|
409
|
|
|
def initPopulation(self, task): |
|
410
|
|
|
r"""Initialize starting population. |
|
411
|
|
|
|
|
412
|
|
|
Args: |
|
413
|
|
|
task (Task): Optimization task. |
|
414
|
|
|
|
|
415
|
|
|
Returns: |
|
416
|
|
|
Tuple[numpy.ndarray, numpy.ndarray[float], Dict[str, Any]]: |
|
417
|
|
|
1. Initialized population. |
|
418
|
|
|
2. Initialized populations function/fitness value. |
|
419
|
|
|
3. Additional arguments: |
|
420
|
|
|
* enable (numpy.ndarray[bool]): If solution/individual is enabled. |
|
421
|
|
|
* improve (numpy.ndarray[bool]): If solution/individual is improved. |
|
422
|
|
|
* SR (numpy.ndarray): Search range. |
|
423
|
|
|
* grades (numpy.ndarray[int]): Grade of solution/individual. |
|
424
|
|
|
""" |
|
425
|
|
|
X, X_f, d = Algorithm.initPopulation(self, task) |
|
426
|
|
|
enable, improve, SR, grades = full(self.NP, True), full(self.NP, True), full([self.NP, task.D], task.bRange / 2), full(self.NP, 0.0) |
|
427
|
|
|
d.update({ |
|
428
|
|
|
'enable': enable, 'improve': improve, 'SR': SR, 'grades': grades |
|
429
|
|
|
}) |
|
430
|
|
|
return X, X_f, d |
|
431
|
|
|
|
|
432
|
|
|
def runIteration(self, task, X, X_f, xb, xb_f, enable, improve, SR, grades, **dparams): |
|
433
|
|
|
r"""Core function of MultipleTrajectorySearch algorithm. |
|
434
|
|
|
|
|
435
|
|
|
Args: |
|
436
|
|
|
task (Task): Optimization task. |
|
437
|
|
|
X (numpy.ndarray): Current population of individuals. |
|
438
|
|
|
X_f (numpy.ndarray[float]): Current individuals function/fitness values. |
|
439
|
|
|
xb (numpy.ndarray): Global best individual. |
|
440
|
|
|
xb_f (float): Global best individual function/fitness value. |
|
441
|
|
|
enable (numpy.ndarray[bool]): Enabled status of individuals. |
|
442
|
|
|
improve (numpy.ndarray[bool]): Improved status of individuals. |
|
443
|
|
|
SR (numpy.ndarray): Search ranges of individuals. |
|
444
|
|
|
grades (numpy.ndarray[int]): Grades of individuals. |
|
445
|
|
|
**dparams (Dict[str, Any]): Additional arguments. |
|
446
|
|
|
|
|
447
|
|
|
Returns: |
|
448
|
|
|
Tuple[numpy.ndarray, numpy.ndarray[float], Dict[str, Any]]: |
|
449
|
|
|
1. Initialized population. |
|
450
|
|
|
2. Initialized populations function/fitness value. |
|
451
|
|
|
3. Additional arguments: |
|
452
|
|
|
* enable (numpy.ndarray[bool]): If solution/individual is enabled. |
|
453
|
|
|
* improve (numpy.ndarray[bool]): If solution/individual is improved. |
|
454
|
|
|
* SR (numpy.ndarray): Search range. |
|
455
|
|
|
* grades (numpy.ndarray[int]): Grade of solution/individual. |
|
456
|
|
|
""" |
|
457
|
|
|
for i in range(len(X)): |
|
458
|
|
|
if not enable[i]: continue |
|
459
|
|
|
enable[i], grades[i] = False, 0 |
|
460
|
|
|
X[i], X_f[i], xb, xb_f, k = self.GradingRun(X[i], X_f[i], xb, xb_f, improve[i], SR[i], task) |
|
461
|
|
|
X[i], X_f[i], xb, xb_f, improve[i], SR[i], grades[i] = self.LsRun(k, X[i], X_f[i], xb, xb_f, improve[i], SR[i], grades[i], task) |
|
462
|
|
|
for _ in range(self.NoLsBest): _, _, xb, xb_f, _, _, _ = MTS_LS1(xb, xb_f, xb, xb_f, False, task.bRange, task, rnd=self.Rand) |
|
463
|
|
|
enable[argsort(grades)[:self.NoEnabled]] = True |
|
|
|
|
|
|
464
|
|
|
return X, X_f, {'enable': enable, 'improve': improve, 'SR': SR, 'grades': grades} |
|
465
|
|
|
|
|
466
|
|
|
class MultipleTrajectorySearchV1(MultipleTrajectorySearch): |
|
467
|
|
|
r"""Implementation of Multiple trajectory search. |
|
468
|
|
|
|
|
469
|
|
|
Algorithm: |
|
470
|
|
|
Multiple trajectory search |
|
471
|
|
|
|
|
472
|
|
|
Date: |
|
473
|
|
|
2018 |
|
474
|
|
|
|
|
475
|
|
|
Authors: |
|
476
|
|
|
Klemen Berkovic |
|
477
|
|
|
|
|
478
|
|
|
License: |
|
479
|
|
|
MIT |
|
480
|
|
|
|
|
481
|
|
|
Reference URL: |
|
482
|
|
|
https://ieeexplore.ieee.org/document/4983179/ |
|
483
|
|
|
|
|
484
|
|
|
Reference paper: |
|
485
|
|
|
Tseng, Lin-Yu, and Chun Chen. "Multiple trajectory search for unconstrained/constrained multi-objective optimization." Evolutionary Computation, 2009. CEC'09. IEEE Congress on. IEEE, 2009. |
|
486
|
|
|
|
|
487
|
|
|
Attributes: |
|
488
|
|
|
Name (List[str]): List of strings representing algorithm name. |
|
489
|
|
|
|
|
490
|
|
|
See Also: |
|
491
|
|
|
* :class:`NiaPy.algorithms.other.MultipleTrajectorySearch`` |
|
492
|
|
|
""" |
|
493
|
|
|
Name = ['MultipleTrajectorySearchV1', 'MTSv1'] |
|
494
|
|
|
|
|
495
|
|
|
@staticmethod |
|
496
|
|
|
def algorithmInfo(): |
|
497
|
|
|
r"""Get basic information of algorithm. |
|
498
|
|
|
|
|
499
|
|
|
Returns: |
|
500
|
|
|
str: Basic information of algorithm. |
|
501
|
|
|
|
|
502
|
|
|
See Also: |
|
503
|
|
|
* :func:`NiaPy.algorithms.Algorithm.algorithmInfo` |
|
504
|
|
|
""" |
|
505
|
|
|
return r"""Tseng, Lin-Yu, and Chun Chen. "Multiple trajectory search for unconstrained/constrained multi-objective optimization." Evolutionary Computation, 2009. CEC'09. IEEE Congress on. IEEE, 2009.""" |
|
506
|
|
|
|
|
507
|
|
|
def setParameters(self, **kwargs): |
|
508
|
|
|
r"""Set core parameters of MultipleTrajectorySearchV1 algorithm. |
|
509
|
|
|
|
|
510
|
|
|
Args: |
|
511
|
|
|
**kwargs (Dict[str, Any]): Additional arguments. |
|
512
|
|
|
|
|
513
|
|
|
See Also: |
|
514
|
|
|
* :func:`NiaPy.algorithms.other.MultipleTrajectorySearch.setParameters` |
|
515
|
|
|
""" |
|
516
|
|
|
kwargs.pop('NoLsBest', None) |
|
517
|
|
|
MultipleTrajectorySearch.setParameters(self, NoLsBest=0, LSs=(MTS_LS1v1, MTS_LS2, MTS_LS3v1), **kwargs) |
|
518
|
|
|
|
|
519
|
|
|
# vim: tabstop=3 noexpandtab shiftwidth=3 softtabstop=3 |
|
520
|
|
|
|
NiaOrg /
NiaPy
Loading history...