-
Notifications
You must be signed in to change notification settings - Fork 1
/
Result.py
166 lines (144 loc) · 8.39 KB
/
Result.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
import json
import os
from statistics import mean,stdev
import numpy as np
class Result:
def __init__(self,input_path,parameters=None):
if parameters is None:
params=input_path.split('/')[-2].split('_')
params[0]=int(params[0][1:])
params[1]=float('0.'+params[1][1:])
self.parameters={'n':params[0],'alpha':params[1]}
exps=dict()
it = 0
for file in os.listdir(input_path):
with open(input_path + file) as f:
for line in f:
l = json.loads(line)
exps[it] = l
it+=1
iterations=len(exps)
methods = [m for m in exps[0]]
models = [m for m in exps[0][methods[0]]]
results={method:{model:[] for model in models} for method in methods}
for method in methods:
for model in models:
for i in range(iterations):
results[method][model].append(exps[i][method][model])
self.results=results
self.iterations=iterations
self.models=models
self.methods=methods
else:
filepath=input_path+'n'+str(parameters['n'])+'_a'+str(parameters['alpha'])+'.json'
with open(filepath) as f:
results=json.load(f)
self.results=results['results']
self.parameters=parameters
self.methods=[method for method in self.results]
self.models=[model for model in self.results[self.methods[0]]]
self.iterations=len(self.results[self.methods[0]][self.models[0]])
def store_summary(self,folder):
path=folder+'n'+str(self.parameters['n'])+'_a'+str(self.parameters['alpha'])+'.json'
stor={'n':self.parameters['n'],'alpha':self.parameters['alpha'],'results':self.results}
with open(path,'w') as f:
json.dump(stor,f)
def rankset_low_up_size(self):
self.lows = {method:{model: [] for model in self.models} for method in self.methods}
self.ups = {method:{model: [] for model in self.models} for method in self.methods}
self.sizes = {method:{model: [] for model in self.models} for method in self.methods}
for method in self.methods:
for model in self.results[method]:
self.lows[method][model] = [s[0] for s in self.results[method][model]]
self.ups[method][model] = [s[1] for s in self.results[method][model]]
self.sizes[method][model] = [s[1] - s[0] + 1 for s in self.results[method][model]]
def set_model_order(self):
models=self.models
avg_lows=[mean(self.lows['baseline'][model]) for model in models]
sorted_indices = np.argsort(avg_lows)
self.models = [models[i] for i in sorted_indices]
def rankset_stats(self):
self.avg_lows={method:[mean(self.lows[method][model]) for model in self.models] for method in self.methods}
self.avg_ups = {method: [mean(self.ups[method][model]) for model in self.models] for method in self.methods}
self.avg_sizes_model = {method: [mean(self.sizes[method][model]) for model in self.models] for method in self.methods}
self.ci_lows = {method: [stdev(self.lows[method][model])*1.96/np.sqrt(self.iterations) for model in self.models] for method in self.methods}
self.ci_ups = {method: [stdev(self.ups[method][model])*1.96/np.sqrt(self.iterations) for model in self.models] for method in self.methods}
self.ci_sizes_model = {method: [stdev(self.sizes[method][model])*1.96/np.sqrt(self.iterations) for model in self.models] for method in self.methods}
all_sizes={method:[self.sizes[method][model] for model in self.models] for method in self.methods}
all_sizes={method:[x for xs in [all_sizes[method][m] for m in range(len(self.models))] for x in xs] for method in self.methods}
self.avg_sizes={method:mean(all_sizes[method]) for method in self.methods}
self.ci_sizes={method:stdev(all_sizes[method])*1.96/np.sqrt(len(all_sizes[method])) for method in self.methods}
def find_ranks(self):
ranks = {method:{model:[0 for _ in range(len(self.models))] for model in self.models} for method in self.methods}
sizes_stats = {method: {model: [0 for _ in range(len(self.models))] for model in self.models} for method in self.methods}
for method in self.methods:
for model in self.models:
for i in range(self.iterations):
for r in range(len(self.models)):
if self.lows[method][model][i]<=r+1 and self.ups[method][model][i]>=r+1:
ranks[method][model][r]+=1
if self.sizes[method][model][i]==r+1:
sizes_stats[method][model][r]+=1
for r in range(len(self.models)):
ranks[method][model][r]/=self.iterations
sizes_stats[method][model][r] /= self.iterations
self.ranks=ranks
self.top_ranks = {method: {model: np.argmax(self.ranks[method][model]) for model in self.models} for method in
self.methods}
self.sizes_stats = sizes_stats
def all_ranksets(self,method,model):
ranksets=dict()
for i in range(self.iterations):
rankset=(self.lows[method][model][i],self.ups[method][model][i])
if rankset in ranksets:
ranksets[rankset]+=1
else:
ranksets[rankset]=1
return ranksets
def total_solutions(self):
self.sols = {method: [1 for _ in range(self.iterations)] for method in self.methods}
for method in self.methods:
for model in self.models:
for i in range(self.iterations):
self.sols[method][i] *= self.sizes[method][model][i]
self.avg_sols = {method: mean(self.sols[method]) for method in self.methods}
self.ci_sols = {method: stdev(self.sols[method])*1.96/np.sqrt(self.iterations) for method in self.methods}
def find_errors(self,type='intersect'):
errors_per_model={method:{model:0 for model in self.models} for method in self.methods if method!='baseline'}
errors_per_iteration={method:[0 for _ in range(self.iterations)] for method in self.methods if method!='baseline'}
correct_iterations={method:[0 for _ in range(self.iterations)] for method in self.methods if method!='baseline'}
singletons={model:[] for model in self.models}
for method in self.methods:
if method=='baseline':
continue
for i in range(self.iterations):
correct_iteration=True
for model in self.models:
if self.lows['baseline'][model][i]==self.ups['baseline'][model][i]:
singletons[model].append(i)
if type=='intersect':
if self.lows[method][model][i]>self.ups['baseline'][model][i] or self.ups[method][model][i]<self.lows['baseline'][model][i]:
errors_per_model[method][model]+=1
errors_per_iteration[method][i]+=1
correct_iteration=False
elif type=='contain':
if self.lows[method][model][i]>self.lows['baseline'][model][i] or self.ups[method][model][i]<self.ups['baseline'][model][i]:
errors_per_model[method][model]+=1
errors_per_iteration[method][i]+=1
correct_iteration=False
if correct_iteration:
correct_iterations[method][i]=1
return singletons, errors_per_model, errors_per_iteration, correct_iterations
def find_correct(self):
_,_,_, correct_iterations_intersect=self.find_errors()
_, _, _, correct_iterations_contain = self.find_errors(type='contain')
self.correct_intersect={method:sum(correct_iterations_intersect[method]) for method in self.methods if method!='baseline'}
self.correct_contain = {method: sum(correct_iterations_contain[method]) for method in self.methods if
method != 'baseline'}
def do_analysis(self):
self.rankset_low_up_size()
self.set_model_order()
self.rankset_stats()
self.total_solutions()
self.find_ranks()
self.find_correct()