forked from justinsvegliato/online-performance-prediction
-
Notifications
You must be signed in to change notification settings - Fork 0
/
computation.py
123 lines (78 loc) · 5 KB
/
computation.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
import copy
import numpy as np
import utils
STOP_SYMBOL = 'STOP'
CONTINUE_SYMBOL = 'CONTINUE'
def get_intrinsic_values(qualities, multiplier):
return np.multiply(multiplier, qualities)
def get_time_costs(steps, multiplier):
return np.exp(np.multiply(multiplier, steps))
def get_comprehensive_values(instrinsic_values, time_costs):
return instrinsic_values - time_costs
def get_mevc(estimated_quality, step, profile_1, profile_3, config):
origin_class = utils.digitize(estimated_quality, config['solution_quality_class_bounds'])
current_expected_value = 0
next_expected_value = 0
for target_class in config['solution_quality_classes']:
target_quality = utils.get_bin_value(target_class, config['solution_quality_class_count'])
intrinsic_value = get_intrinsic_values(target_quality, config['intrinsic_value_multiplier'])
current_time_cost = get_time_costs(step, config['time_cost_multiplier'])
current_comprehensive_value = get_comprehensive_values(intrinsic_value, current_time_cost)
current_expected_value += profile_3[origin_class][step][target_class] * current_comprehensive_value
next_time_cost = get_time_costs(step + 1, config['time_cost_multiplier'])
next_comprehensive_value = get_comprehensive_values(intrinsic_value, next_time_cost)
next_expected_value += profile_1[origin_class][step][target_class] * next_comprehensive_value
return next_expected_value - current_expected_value
def get_optimal_values(profile_2, profile_3, config, epsilon=0.2):
limit = len(profile_3[0])
values = {origin_class: limit * [0] for origin_class in config['solution_quality_classes']}
while True:
new_values = copy.deepcopy(values)
delta = 0
for origin_class in config['solution_quality_classes']:
for step in range(limit):
if step + 1 < limit:
stop_value = 0
continue_value = 0
for target_class in config['solution_quality_classes']:
target_quality = utils.get_bin_value(target_class, config['solution_quality_class_count'])
intrinsic_value = get_intrinsic_values(target_quality, config['intrinsic_value_multiplier'])
time_cost = get_time_costs(step, config['time_cost_multiplier'])
comprehensive_value = get_comprehensive_values(intrinsic_value, time_cost)
stop_value += profile_3[origin_class][step][target_class] * comprehensive_value
continue_value += profile_2[origin_class][step][target_class] * values[target_class][step + 1]
new_values[origin_class][step] = max(stop_value, continue_value)
delta = max(delta, abs(new_values[origin_class][step] - values[origin_class][step]))
values = new_values
print('Delta: %f' % delta)
if delta < epsilon:
return values
def get_optimal_action(quality, step, values, profile_2, profile_3, config):
origin_class = utils.digitize(quality, config['solution_quality_class_bounds'])
stop_value = 0
continue_value = 0
for target_class in config['solution_quality_classes']:
target_quality = utils.get_bin_value(target_class, config['solution_quality_class_count'])
intrinsic_value = get_intrinsic_values(target_quality, config['intrinsic_value_multiplier'])
time_cost = get_time_costs(step, config['time_cost_multiplier'])
comprehensive_value = get_comprehensive_values(intrinsic_value, time_cost)
stop_value += profile_3[origin_class][step][target_class] * comprehensive_value
continue_value += profile_2[origin_class][step][target_class] * values[target_class][step + 1]
return STOP_SYMBOL if stop_value >= continue_value else CONTINUE_SYMBOL
def get_policy(values, profile_2, profile_3, config):
limit = len(profile_3[0])
policy = {quality_class: limit * [0] for quality_class in config['solution_quality_classes']}
for quality_class in config['solution_quality_classes']:
for step in range(limit):
if step + 1 < limit:
stop_value = 0
continue_value = 0
for target_class in config['solution_quality_classes']:
target_quality = utils.get_bin_value(target_class, config['solution_quality_class_count'])
intrinsic_value = get_intrinsic_values(target_quality, config['intrinsic_value_multiplier'])
time_cost = get_time_costs(step, config['time_cost_multiplier'])
comprehensive_value = get_comprehensive_values(intrinsic_value, time_cost)
stop_value += profile_3[quality_class][step][target_class] * comprehensive_value
continue_value += profile_2[quality_class][step][target_class] * values[target_class][step + 1]
policy[quality_class][step] = STOP_SYMBOL if stop_value >= continue_value else CONTINUE_SYMBOL
return policy