-
Notifications
You must be signed in to change notification settings - Fork 1
/
optimizer.py
163 lines (144 loc) · 7.58 KB
/
optimizer.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
#!/usr/bin/env python
import time
import pygmo as pg
from minimization_problem import MinimizationProblem
from optimizers.boa import BOA
from optimizers.saboa import SABOA
from optimizers.xboa import XBOA
class Optimizer:
"""
Solve the problem using pygmo algorithms
"""
def __init__(self, problem, method='random', pop_size=10, iterations=1,
verbosity_level=None, seed=None, early_stopping_counter=10, **kwargs):
"""
Creates an Optimizer object for solving the problem
Arguments:
- problem: A pygmo problem object to optimize
- method: optimization algorithm name
- pop_size: number of solutions in the population (ignored when method is 'random')
- iterations: number of generations
- verbosity_level: set log frequecy for pygmo optimizers
- seed: set a fixed seed for generating stochastic variables
- early_stopping_counter: number of iterations with no fitness improvemnt to stop
- **kwargs: captures additional args for tuning the solver (depend on the method)
"""
self.interations = iterations
self.early_stopping_counter = early_stopping_counter
# create the population
self.population = pg.population(problem, pop_size, seed=seed)
# create the solver
if method.lower() in ['boa', 'mboa', 'aboa']:
self.solver = self.boa_solver(max_gen=iterations, variant=method, **kwargs)
self.custom_algorithm = True
elif method.lower() == 'saboa':
self.solver = self.saboa_solver(max_gen=iterations, **kwargs)
self.custom_algorithm = True
elif method.lower() in ['xboa', 'xaboa']:
self.solver = self.xboa_solver(max_gen=iterations, variant=method, **kwargs)
self.custom_algorithm = True
else:
self.solver = 'random'
self.solver_name = 'Random'
# set verbosity level
if verbosity_level is not None and verbosity_level > 0:
self.verbosity_level = verbosity_level
if self.solver == 'random':
print("Solving problem using: Random")
else:
print("Solving problem using {}".format(self.solver.get_name()))
else:
self.verbosity_level = 0
def get_solution(self):
"""
Launch the optimization process and return the best solution found
"""
if self.solver == 'random':
best_solution = self.population.get_x()[0]
else:
early_stopping_counter = self.early_stopping_counter
old_best_fitness = self.population.champion_f[0]
if self.verbosity_level > 0:
print("{:^7}{:^10}{:>10}{:>10}{:>10}".format(
"Gen", "Fevals", "Fbest", "Improv", "Duration"))
for i in range(1, self.interations+1):
# evolve the population for 1 iteration at a time
time_start = time.time()
alg = self.solver.extract(object)
if self.custom_algorithm:
alg.set_iter(i)
self.population = self.solver.evolve(self.population)
duration = round(time.time() - time_start, 2)
# get best fitness so far and save statistics
best_fitness = self.population.champion_f[0]
fitness_evaluations = self.population.problem.get_fevals()
# check early stopping condition
improvemnt = best_fitness - old_best_fitness
if improvemnt == 0:
early_stopping_counter -= 1
else:
early_stopping_counter = self.early_stopping_counter
old_best_fitness = best_fitness
# log statistics on screen
if self.verbosity_level > 0 and i % self.verbosity_level == 0:
print("{:^7}{:^10}{:>10}{:>10}{:>10}".format(i,
fitness_evaluations,
round(best_fitness, 6),
round(improvemnt, 6),
duration))
# early stopping
if early_stopping_counter <= 0:
break
if self.verbosity_level > 0:
early_stop_msg = "(early stopping)" if early_stopping_counter <= 0 else ""
print("Executed: {} generations {}".format(i, early_stop_msg))
best_solution = self.population.champion_x
return best_solution
def boa_solver(self, sensory_modality=0.01, power_exponent=0.1,
switch_probability=0.8, mu=2, variant="BOA", max_gen=1, **unused_args):
"""
Solve the problem using Burtterfly Optimization Algorithm.
Arguments:
- sensory_modality: regulates the degree of absorption of the fragrance
- power_exponent: regulates the response's sensitivity to sensory stimulus
- switch_probability: used to switch between global search and local search
- mu: used for ABOA variant only (for more information see [Zhang et al. 2020])
- variant: Select which variant to execute from the following cases:
1. "BOA": The standard BOA algorithm from [Arora et al. 2019] & [Arora et al. 2016]
2. "mBOA": BOA with intensive search from [Arora et al. 2018]
3. "ABOA": BOA with a non-linear update rule for the sensory modality from [Zhang et al. 2020]
- max_gen: max number of generations (used for updating the sensory modality)
- **unused_args: capture additional args that may have been added in Optimizer config
"""
solver = pg.algorithm(BOA(gen=1, c=sensory_modality, a=power_exponent,
p=switch_probability, mu=mu, variant=variant, max_gen=max_gen))
return solver
def saboa_solver(self, switch_probability=0.8, max_gen=1, **unused_args):
"""
Solve the problem using Self-Adaptative Burtterfly Optimization Algorithm.
[Fan et al. 2020]
Arguments:
- switch_probability: used to switch between global search and local search
- max_gen: max number of generations (used for updating the sensory modality)
- **unused_args: capture additional args that may have been added in Optimizer config
"""
solver = pg.algorithm(SABOA(gen=1, p=switch_probability, max_gen=max_gen))
return solver
def xboa_solver(self, sensory_modality=0.01, power_exponent=0.1,
switch_probability=0.8, mu=2, variant="xBOA", max_gen=1, **unused_args):
"""
Solve the problem using Crossover Burtterfly Optimization Algorithm.
Arguments:
- sensory_modality: regulates the degree of absorption of the fragrance
- power_exponent: regulates the response's sensitivity to sensory stimulus
- switch_probability: used to switch between global search and local search
- mu: Used for xABOA variant only (for more information see [Zhang et al. 2020])
- variant: Select which variant to execute from the following cases:
1. "xBOA": The xBOA algorithm from [Bendahmane et al. 2021]
2. "xABOA": xBOA + the non-linear update rule for the sensory modality from [Zhang et al. 2020]
- max_gen: max number of generations (used for updating the sensory modality)
- **unused_args: capture additional args that may have been added in Optimizer config
"""
solver = pg.algorithm(XBOA(gen=1, c=sensory_modality, a=power_exponent,
p=switch_probability, mu=mu, variant=variant, max_gen=max_gen))
return solver