|
1
|
|
|
# Author: Simon Blanke |
|
2
|
|
|
# Email: [email protected] |
|
3
|
|
|
# License: MIT License |
|
4
|
|
|
|
|
5
|
|
|
import numpy as np |
|
6
|
|
|
from tqdm import tqdm |
|
7
|
|
|
|
|
8
|
|
|
|
|
9
|
|
|
class ProgressBarBase: |
|
10
|
|
|
def __init__(self, nth_process, n_iter, objective_function): |
|
11
|
|
|
self._score_best = -np.inf |
|
12
|
|
|
self.score_best_list = [] |
|
13
|
|
|
|
|
14
|
|
|
self.convergence_data = [] |
|
15
|
|
|
|
|
16
|
|
|
self._best_since_iter = 0 |
|
17
|
|
|
self.best_since_iter_list = [] |
|
18
|
|
|
|
|
19
|
|
|
self.objective_function = objective_function |
|
20
|
|
|
|
|
21
|
|
|
self.n_iter_current = 0 |
|
22
|
|
|
|
|
23
|
|
|
@property |
|
24
|
|
|
def score_best(self): |
|
25
|
|
|
return self._score_best |
|
26
|
|
|
|
|
27
|
|
|
@score_best.setter |
|
28
|
|
|
def score_best(self, score): |
|
29
|
|
|
self.score_best_list.append(score) |
|
30
|
|
|
self._score_best = score |
|
31
|
|
|
|
|
32
|
|
|
@property |
|
33
|
|
|
def best_since_iter(self): |
|
34
|
|
|
return self._best_since_iter |
|
35
|
|
|
|
|
36
|
|
|
@best_since_iter.setter |
|
37
|
|
|
def best_since_iter(self, nth_iter): |
|
38
|
|
|
self.best_since_iter_list.append(nth_iter) |
|
39
|
|
|
self._best_since_iter = nth_iter |
|
40
|
|
|
|
|
41
|
|
|
def _new2best(self, score_new, pos_new, nth_iter): |
|
42
|
|
|
if score_new > self.score_best: |
|
43
|
|
|
self.score_best = score_new |
|
44
|
|
|
self.pos_best = pos_new |
|
45
|
|
|
self.best_since_iter = nth_iter |
|
46
|
|
|
|
|
47
|
|
|
self.convergence_data.append(self.score_best) |
|
48
|
|
|
|
|
49
|
|
|
|
|
50
|
|
|
class ProgressBarLVL0(ProgressBarBase): |
|
51
|
|
|
def __init__(self, nth_process, n_iter, objective_function): |
|
52
|
|
|
super().__init__(nth_process, n_iter, objective_function) |
|
53
|
|
|
|
|
54
|
|
|
def update(self, score_new, pos_new, nth_iter): |
|
55
|
|
|
self.n_iter_current = nth_iter |
|
56
|
|
|
self._new2best(score_new, pos_new, nth_iter) |
|
57
|
|
|
|
|
58
|
|
|
def close(self): |
|
59
|
|
|
pass |
|
60
|
|
|
|
|
61
|
|
|
|
|
62
|
|
|
class ProgressBarLVL1(ProgressBarBase): |
|
63
|
|
|
def __init__(self, nth_process, n_iter, objective_function): |
|
64
|
|
|
super().__init__(nth_process, n_iter, objective_function) |
|
65
|
|
|
self._tqdm = tqdm( |
|
66
|
|
|
**self._tqdm_dict(nth_process, n_iter, objective_function) |
|
67
|
|
|
) |
|
68
|
|
|
|
|
69
|
|
|
def update(self, score_new, pos_new, nth_iter): |
|
70
|
|
|
self.n_iter_current = nth_iter |
|
71
|
|
|
|
|
72
|
|
|
self._new2best(score_new, pos_new, nth_iter) |
|
73
|
|
|
|
|
74
|
|
|
if score_new > self.score_best: |
|
75
|
|
|
self._tqdm.set_postfix( |
|
76
|
|
|
best_score=str(score_new), |
|
77
|
|
|
best_pos=str(pos_new), |
|
78
|
|
|
best_iter=str(self._best_since_iter), |
|
79
|
|
|
) |
|
80
|
|
|
|
|
81
|
|
|
self._tqdm.update(1) |
|
82
|
|
|
# self._tqdm.refresh() |
|
83
|
|
|
|
|
84
|
|
|
def close(self): |
|
85
|
|
|
self._tqdm.close() |
|
86
|
|
|
|
|
87
|
|
|
def _tqdm_dict(self, nth_process, n_iter, objective_function): |
|
88
|
|
|
"""Generates the parameter dict for tqdm in the iteration-loop of each optimizer""" |
|
89
|
|
|
|
|
90
|
|
|
self.objective_function = objective_function |
|
91
|
|
|
|
|
92
|
|
|
if nth_process is None: |
|
93
|
|
|
process_str = "" |
|
94
|
|
|
else: |
|
95
|
|
|
process_str = "Process " + str(nth_process) |
|
96
|
|
|
|
|
97
|
|
|
return { |
|
98
|
|
|
"total": n_iter, |
|
99
|
|
|
"desc": process_str, |
|
100
|
|
|
"position": nth_process, |
|
101
|
|
|
"leave": False, |
|
102
|
|
|
# "smoothing": 1.0, |
|
103
|
|
|
} |
|
104
|
|
|
|
|
105
|
|
|
|
SimonBlanke /
Gradient-Free-Optimizers
