forked from cmudrc/SBCE-simulator
-
Notifications
You must be signed in to change notification settings - Fork 0
/
design_specialist.py
176 lines (126 loc) · 6.13 KB
/
design_specialist.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
164
165
166
167
168
169
170
171
172
#!/usr/bin/env python
# coding: utf-8
# In[1]:
import numpy as np
import random
import copy
class design_specialist:
#hidden markov init
#def __init__(self,num_states,num_ops,learn_rate):
def __init__(self,num_ops,learn_rate,temp,num_designs):
self.learn_rate = learn_rate;
self.selected_design = 0
self.requested_by = []
#objective function:
#actions:
self.move_ids = np.arange(0,num_ops)
self.num_designs = num_designs
#Learning mechanics: Markov
self.move_prob = np.ones((num_ops,num_ops))/num_ops
self.move = np.random.choice(self.move_ids)
#Learning mechanics: hidden markov
#self.state_ids = np.arange(0,num_states)
#self.trans_prob = np.zeros(num_states,num_states)
#self.op_prefs = np.zeros(num_states,num_ops)
#self.hidden_state = random_choice(self.state_ids)
#Triki temp stuff
self.t_init = temp
self.delt_init = 1e6
self.hist_init = []
self.temp = np.ones(num_designs)*self.t_init
self.delt = np.ones(num_designs)*self.delt_init
self.hist_length = 10;
self.active_temp = self.t_init
self.active_delt = self.delt_init
self.active_last_move = random.randint(0,num_ops-1)
self.move = random.randint(0,num_ops-1)
self.last_moves = np.random.randint(0,num_ops,num_designs)
self.hist = [copy.deepcopy(self.hist_init) for i in range(num_designs)]
self.hist_needs = [copy.deepcopy(self.hist_init) for i in range(num_designs)]
self.hist_target_needs = [copy.deepcopy(self.hist_init) for i in range(num_designs)]
self.hist_designs = [copy.deepcopy(self.hist_init) for i in range(num_designs)]
self.hist_solutions = [copy.deepcopy(self.hist_init) for i in range(num_designs)]
self.active_hist = copy.deepcopy(self.hist_init)
self.active_hist_needs = copy.deepcopy(self.hist_init)
self.active_hist_target_needs = copy.deepcopy(self.hist_init)
self.active_hist_design = copy.deepcopy(self.hist_init)
self.active_hist_solution = copy.deepcopy(self.hist_init)
def iterate(self):
pass
def select_move(self):
#Markov learning
#pick a new move # save it
self.active_last_move = self.move
self.move = random.choices(self.move_ids,self.move_prob[self.active_last_move])[0]
#HIDDEN MARKOV
#store current statistics
#self.old_state = self.hidden_state
#pick new state
#self.hidden_state = random.choices(self.move_ids,self.trans_prob[self.hidden_state])
#pick a new move
#self.move = random.choices(self.move_ids,self.op_prefs[self.hidden_state])
#Move
return self.move
def update_learn(self, d_qual):
if d_qual <= 0:
#markov
self.move_prob[self.active_last_move][self.move] *= (1+self.learn_rate);
accept = 1
#hidden markov
#self.op_prefs[self.hidden_state,self.move] *= (1+self.learn_rate);
#self.trans_prob[self.old_state,self.hidden_state] *= (1+self.learn_rate);
elif d_qual > 0:
#markov
self.move_prob[self.active_last_move][self.move] *= (1-self.learn_rate);
if random.uniform(0,1) < np.exp(-d_qual/self.active_temp):
accept = 1
else:
accept = 0
#hidden markov
#self.op_prefs[self.hidden_state,self.move] *= (1-self.learn_rate);
#self.trans_prob[self.hidden_state,self.move] *= (1-self.learn_rate);
#Normalize arrays
#markov
self.move_prob[self.active_last_move] /= np.sum(self.move_prob[self.active_last_move])
return accept
#hidden markov
#self.op_prefs /= np.sum(self.op_prefs)
#self.trans_prob /= np.sum(self.trans_prob)
def update_temp(self):
if len(self.active_hist) <= 1:
return
else:
var = np.var(np.asarray(self.active_hist))
if var > 0:
update_factor = self.active_delt * self.active_temp / var
if update_factor > 1:
self.active_delt /= 2
update_factor /= 2
if update_factor < 1:
self.active_temp = self.active_temp*(1-update_factor)
def reset_temps(self):
self.temp = np.ones(self.num_designs)*self.t_init
self.delt = np.ones(self.num_designs)*self.delt_init
self.active_temp = self.t_init
self.active_delt = self.delt_init
def reset_active_temps(self):
self.active_temp = self.t_init
self.active_delt = self.delt_init
def reset_histories(self):
self.hist = [copy.deepcopy(self.hist_init) for i in range(self.num_designs)]
self.hist_needs = [copy.deepcopy(self.hist_init) for i in range(self.num_designs)]
self.hist_target_needs = [copy.deepcopy(self.hist_init) for i in range(self.num_designs)]
self.hist_designs = [copy.deepcopy(self.hist_init) for i in range(self.num_designs)]
self.hist_solutions = [copy.deepcopy(self.hist_init) for i in range(self.num_designs)]
self.active_hist = copy.deepcopy(self.hist_init)
self.active_hist_needs = copy.deepcopy(self.hist_init)
self.active_hist_target_needs = copy.deepcopy(self.hist_init)
self.active_hist_design = copy.deepcopy(self.hist_init)
self.active_hist_solution = copy.deepcopy(self.hist_init)
def reset_active_histories(self):
self.active_hist = copy.deepcopy(self.hist_init)
self.active_hist_needs = copy.deepcopy(self.hist_init)
self.active_hist_target_needs = copy.deepcopy(self.hist_init)
self.active_hist_design = copy.deepcopy(self.hist_init)
self.active_hist_solution = copy.deepcopy(self.hist_init)
# In[ ]: