-
Notifications
You must be signed in to change notification settings - Fork 5
/
plots_anova_bayesian.py
148 lines (122 loc) · 5.5 KB
/
plots_anova_bayesian.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
import argparse
import matplotlib.pyplot as plt
import seaborn
import pandas as pd
import numpy as np
import pystan
import scipy
import sys
import bayesian_anova
parser = argparse.ArgumentParser(description='ANOVA plot generator')
parser.add_argument('--dataset', dest='dataset', action='store',
choices=['mnist','svhn','cifar10'],
help='Dataset', required=True)
args = parser.parse_args(sys.argv[1:])
dataset = args.dataset
if dataset == 'mnist':
nn_model = 'mlp'
experiments = ['mlp', 'mlp-dropout', 'mlp-poor-bayesian', 'mlp-bayesian']
else:
nn_model = 'convolutional'
experiments = ['convolutional', 'convolutional-dropout', 'convolutional-poor-bayesian']
cols = ['experiment_name',
'test_acc',
'train_time',
'entropy__auc',
'entropy_expectation__auc',
'classifier__auc']
dfs = []
for exp in experiments:
df_with = pd.read_csv(dataset+'_results/'+exp+'_with_unknown.csv')
df_with_results = df_with[cols].set_index('experiment_name')
df_without = pd.read_csv(dataset+'_results/'+exp+'_out_unknown.csv')
df_without_results = df_without[cols].set_index('experiment_name')
dfs.append([exp, df_with_results, df_without_results])
results_cols = ['experiment',
'in_test_acc', 'out_test_acc',
'in_train_time', 'out_train_time',
'in_entropy_auc', 'out_entropy_auc',
'in_entropy_expectation_auc', 'out_entropy_expectation_auc',
'in_classifier_auc', 'out_classifier_auc']
dfs_results = []
for exp, df_with, df_without in dfs:
results = pd.DataFrame(columns=results_cols)
for (in_key, *in_row), (out_key, *out_row) in zip(df_with.itertuples(), df_without.itertuples()):
assert in_key == out_key
results.loc[len(results)] = [
str(in_key),
in_row[0], out_row[0],
in_row[1], out_row[1],
in_row[2], out_row[2],
in_row[3], out_row[3],
in_row[4], out_row[4],
]
dfs_results.append([exp, results])
final_results_cols = ['experiment']
for c in results_cols[1:]:
for exp, _ in dfs_results:
final_results_cols.append(exp+'_'+c)
final_results = pd.DataFrame(columns=final_results_cols)
for key_row in zip(*[df.itertuples() for exp, df in dfs_results]):
for i in range(1, len(key_row)):
_, *prev_row = key_row[i-1]
_, *row = key_row[i]
assert prev_row[0] == row[0]
_, *row = key_row[0]
new_row = [row[0]]
for c in range(1, len(results_cols)):
for _, *row in key_row:
new_row.append(row[c])
final_results.loc[len(final_results)] = new_row
model = pystan.StanModel(model_code=bayesian_anova.one_way_code)
out_acc = [nn_model+'_out_classifier_auc',
nn_model+'-dropout_out_classifier_auc',
nn_model+'-poor-bayesian_out_classifier_auc',
nn_model+'-bayesian_out_classifier_auc']
y_out = final_results[out_acc].values
in_acc = [nn_model+'_in_classifier_auc',
nn_model+'-dropout_in_classifier_auc',
nn_model+'-poor-bayesian_in_classifier_auc',
nn_model+'-bayesian_in_classifier_auc']
y_in = final_results[in_acc].values
model = pystan.StanModel(model_code=bayesian_anova.two_way_code)
N, K = y_out.shape
data = {'K': K, 'N': N, 'y_in': y_in, 'y_out': y_out}
fit = model.sampling(data=data, iter=100000, warmup = 10000, chains=4, thin=5)
bayesian_anova.show_results(fit)
trace = fit.extract()
deterministic = trace['theta'][:,0]
dropout = trace['theta'][:,1]
poor_bayesian = trace['theta'][:,2]
bayesian = trace['theta'][:,3]
in_mean = trace['mu_in']
out_mean = trace['mu_out']
traces = [out_mean, in_mean, deterministic, dropout, poor_bayesian, bayesian]
traces_name = ['Blind Mean', 'Calibrated Mean', 'ML effect', 'BD effect', 'OSBA effect', 'SV effect']
fig_hist, figs_effects = bayesian_anova.plot_traces(traces, traces_name, show=False)
<<<<<<< HEAD
# fig_hist.savefig(dataset+'_results/images/hist.png')
# fig_effects.savefig(dataset+'_results/images/effects.png')
=======
>>>>>>> c5ee8c1f199975dc77d03542a026ed1507016dcb
fig_hist.savefig(dataset+'_results/images/hist.png')
for name, fig in figs_effects:
name = name.lower().replace(' ', '_')
fig.savefig(dataset+'_results/images/'+name+'.png')
<<<<<<< HEAD
=======
>>>>>>> c5ee8c1f199975dc77d03542a026ed1507016dcb
fig_diff_drop_ml = bayesian_anova.effect_difference(dropout, deterministic, 'BD', 'ML', show=False)
fig_diff_os_ml = bayesian_anova.effect_difference(poor_bayesian, deterministic, 'OSBA', 'ML', show=False)
fig_diff_bayesian_ml = bayesian_anova.effect_difference(bayesian, deterministic, 'SV', 'ML', show=False)
fig_diff_os_drop = bayesian_anova.effect_difference(poor_bayesian, dropout, 'OSBA', 'BD', show=False)
fig_diff_os_bayes = bayesian_anova.effect_difference(poor_bayesian, bayesian, 'OSBA', 'SV', show=False)
fig_diff_drop_bayes = bayesian_anova.effect_difference(dropout, bayesian, 'BD', 'SV', show=False)
fig_diff_iou_io = bayesian_anova.effect_difference(in_mean, out_mean, 'Calibrated', 'Blind', show=False)
fig_diff_drop_ml.savefig(dataset+'_results/images/diff_drop_ml.png')
fig_diff_os_ml.savefig(dataset+'_results/images/diff_os_ml.png')
fig_diff_bayesian_ml.savefig(dataset+'_results/images/diff_bayesian_ml.png')
fig_diff_os_drop.savefig(dataset+'_results/images/diff_os_drop.png')
fig_diff_os_bayes.savefig(dataset+'_results/images/diff_os_bayes.png')
fig_diff_drop_bayes.savefig(dataset+'_results/images/diff_drop_bayes.png')
fig_diff_iou_io.savefig(dataset+'_results/images/diff_iou_io.png')