|
1
|
|
|
# Author: Simon Blanke |
|
2
|
|
|
# Email: [email protected] |
|
3
|
|
|
# License: MIT License |
|
4
|
|
|
|
|
5
|
|
|
from ..util import sort_for_best |
|
6
|
|
|
import numpy as np |
|
7
|
|
|
|
|
8
|
|
|
|
|
9
|
|
|
class InitSearchPosition: |
|
10
|
|
|
def __init__(self, space, model, _main_args_): |
|
11
|
|
|
self._space_ = space |
|
12
|
|
|
self._model_ = model |
|
13
|
|
|
self.init_config = _main_args_.init_config |
|
14
|
|
|
self.X = _main_args_.X |
|
15
|
|
|
self.y = _main_args_.y |
|
16
|
|
|
|
|
17
|
|
|
def _warm_start(self): |
|
18
|
|
|
pos = [] |
|
19
|
|
|
|
|
20
|
|
|
for hyperpara_name in self._space_.search_space.keys(): |
|
21
|
|
|
if hyperpara_name not in list(self._space_.init_para.keys()): |
|
22
|
|
|
# print(hyperpara_name, "not in warm_start selecting random scalar") |
|
23
|
|
|
search_position = self._space_.get_random_pos_scalar(hyperpara_name) |
|
24
|
|
|
|
|
25
|
|
|
else: |
|
26
|
|
|
search_position = self._space_.search_space[hyperpara_name].index( |
|
27
|
|
|
self._space_.init_para[hyperpara_name] |
|
28
|
|
|
) |
|
29
|
|
|
pos.append(search_position) |
|
30
|
|
|
|
|
31
|
|
|
return np.array(pos) |
|
32
|
|
|
|
|
33
|
|
|
def _set_start_pos(self): |
|
34
|
|
|
if self._space_.init_type == "warm_start": |
|
35
|
|
|
pos = self._warm_start() |
|
36
|
|
|
elif self._space_.init_type == "scatter_init": |
|
37
|
|
|
pos = self._scatter_init() |
|
38
|
|
|
else: |
|
39
|
|
|
pos = self._space_.get_random_pos() |
|
40
|
|
|
|
|
41
|
|
|
return pos |
|
42
|
|
|
|
|
43
|
|
|
def _warm_start_scatter_init(self, nth_process): |
|
44
|
|
|
if self.n_warm_start_keys > nth_process: |
|
45
|
|
|
pos = self._create_warm_start() |
|
46
|
|
|
else: |
|
47
|
|
|
pos = self._scatter_init() |
|
48
|
|
|
|
|
49
|
|
|
return pos |
|
50
|
|
|
|
|
51
|
|
|
def _scatter_init(self): |
|
52
|
|
|
pos_list = [] |
|
53
|
|
|
for _ in range(self._space_.init_para["scatter_init"]): |
|
54
|
|
|
pos = self._space_.get_random_pos() |
|
55
|
|
|
pos_list.append(pos) |
|
56
|
|
|
|
|
57
|
|
|
pos_best_list, score_best_list = self._scatter_train(pos_list) |
|
58
|
|
|
pos_best_sorted, _ = sort_for_best(pos_best_list, score_best_list) |
|
59
|
|
|
|
|
60
|
|
|
return pos_best_sorted[0] |
|
61
|
|
|
|
|
62
|
|
|
def _scatter_train(self, pos_list): |
|
63
|
|
|
pos_best_list = [] |
|
64
|
|
|
score_best_list = [] |
|
65
|
|
|
|
|
66
|
|
|
X, y = self._get_random_sample(self.X, self.y) |
|
67
|
|
|
|
|
68
|
|
|
for pos in pos_list: |
|
69
|
|
|
para = self._space_.pos2para(pos) |
|
70
|
|
|
score, eval_time, model = self._model_.train_model(para) |
|
71
|
|
|
|
|
72
|
|
|
pos_best_list.append(pos) |
|
73
|
|
|
score_best_list.append(score) |
|
74
|
|
|
|
|
75
|
|
|
return pos_best_list, score_best_list |
|
76
|
|
|
|
|
77
|
|
|
def _get_random_sample(self, X, y): |
|
78
|
|
|
if isinstance(X, np.ndarray) and isinstance(y, np.ndarray): |
|
79
|
|
|
n_samples = int(X.shape[0] / self._space_.init_para["scatter_init"]) |
|
80
|
|
|
|
|
81
|
|
|
idx = np.random.choice(np.arange(len(X)), n_samples, replace=False) |
|
82
|
|
|
|
|
83
|
|
|
X_sample = X[idx] |
|
84
|
|
|
y_sample = y[idx] |
|
85
|
|
|
|
|
86
|
|
|
return X_sample, y_sample |
|
87
|
|
|
|
SimonBlanke /
Hyperactive
