|
1
|
|
|
# Author: Simon Blanke |
|
2
|
|
|
# Email: [email protected] |
|
3
|
|
|
# License: MIT License |
|
4
|
|
|
|
|
5
|
|
|
import random |
|
6
|
|
|
import numpy as np |
|
7
|
|
|
|
|
8
|
|
|
|
|
9
|
|
|
class SearchSpace: |
|
10
|
|
|
def __init__(self, _core_, model_nr): |
|
11
|
|
|
self.search_space = _core_.search_config[list(_core_.search_config)[model_nr]] |
|
12
|
|
|
self.pos_space_limit() |
|
13
|
|
|
|
|
14
|
|
|
self.init_para = _core_.init_config[list(_core_.init_config)[model_nr]] |
|
15
|
|
|
self.memory = {} |
|
16
|
|
|
|
|
17
|
|
|
if list(self.init_para.keys())[0] == list(self.search_space.keys())[0]: |
|
18
|
|
|
self.init_type = "warm_start" |
|
19
|
|
|
elif list(self.init_para.keys())[0] == "scatter_init": |
|
20
|
|
|
self.init_type = "scatter_init" |
|
21
|
|
|
else: |
|
22
|
|
|
self.init_type = None |
|
23
|
|
|
|
|
24
|
|
|
def load_memory(self, para, score): |
|
25
|
|
|
if para is None or score is None: |
|
26
|
|
|
return |
|
27
|
|
|
for idx in range(para.shape[0]): |
|
28
|
|
|
pos = self.para2pos(para.iloc[[idx]]) |
|
29
|
|
|
pos_str = pos.tostring() |
|
30
|
|
|
self.memory[pos_str] = float(score.values[idx]) |
|
31
|
|
|
|
|
32
|
|
|
def pos_space_limit(self): |
|
33
|
|
|
dim = [] |
|
34
|
|
|
|
|
35
|
|
|
for pos_key in self.search_space: |
|
36
|
|
|
dim.append(len(self.search_space[pos_key]) - 1) |
|
37
|
|
|
|
|
38
|
|
|
self.dim = np.array(dim) |
|
39
|
|
|
|
|
40
|
|
|
def get_random_pos(self): |
|
41
|
|
|
pos_new = np.random.uniform(np.zeros(self.dim.shape), self.dim, self.dim.shape) |
|
42
|
|
|
pos = np.rint(pos_new).astype(int) |
|
43
|
|
|
|
|
44
|
|
|
return pos |
|
45
|
|
|
|
|
46
|
|
|
def get_random_pos_scalar(self, hyperpara_name): |
|
47
|
|
|
n_para_values = len(self.search_space[hyperpara_name]) |
|
48
|
|
|
pos = random.randint(0, n_para_values - 1) |
|
49
|
|
|
|
|
50
|
|
|
return pos |
|
51
|
|
|
|
|
52
|
|
|
def para2pos(self, para): |
|
53
|
|
|
pos_list = [] |
|
54
|
|
|
|
|
55
|
|
|
for pos_key in self.search_space: |
|
56
|
|
|
value = para[[pos_key]].values |
|
57
|
|
|
|
|
58
|
|
|
pos = self.search_space[pos_key].index(value) |
|
59
|
|
|
pos_list.append(pos) |
|
60
|
|
|
|
|
61
|
|
|
return np.array(pos_list) |
|
62
|
|
|
|
|
63
|
|
|
def pos2para(self, pos): |
|
64
|
|
|
if len(self.search_space.keys()) == pos.size: |
|
65
|
|
|
values_dict = {} |
|
66
|
|
|
for i, key in enumerate(self.search_space.keys()): |
|
67
|
|
|
pos_ = int(pos[i]) |
|
68
|
|
|
values_dict[key] = list(self.search_space[key])[pos_] |
|
69
|
|
|
|
|
70
|
|
|
return values_dict |
|
71
|
|
|
|
SimonBlanke /
Hyperactive
