forked from blafabregue/TimeSeriesDeepClustering
-
Notifications
You must be signed in to change notification settings - Fork 0
/
extract_stats.py
157 lines (136 loc) · 6.66 KB
/
extract_stats.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
"""
Script to extract stats into csv files
Author:
Baptiste Lafabregue 2019.25.04
"""
import os
import pathlib
import argparse
import pandas as pd
import numpy as np
import utils
ELEMENT_COUNT = 6
def get_elements(file_name):
split = file_name.split('_')
arch = split[0]
limit = ELEMENT_COUNT
shift = 0
if arch == "dilated" or arch == "bi" or arch == 'res':
arch += "_" + split[1]
enc_loss = split[2]
limit += 1
else:
enc_loss = split[1]
if len(split) < limit:
clust_loss = "None"
shift = 1
else:
clust_loss = split[-2]
dropout = split[(shift-4)]
noise = split[(shift-3)]
dataset_name = split[-1]
return arch, enc_loss, clust_loss, dataset_name, dropout, noise
def main(itr, seed_itrs, type, stat_path, root_dir):
dfs = []
for seed in seed_itrs:
stats = []
stat_path = root_dir + '/stats/' + str(itr) + '/' + str(seed)
stats_files = [name for name in os.listdir(stat_path)
if os.path.isfile(os.path.join(stat_path, name)) and
pathlib.Path(name).suffix != '.error']
stats_files.sort()
error_files = [name for name in os.listdir(stat_path)
if os.path.isfile(os.path.join(stat_path, name)) and
pathlib.Path(name).suffix == '.error']
error_files.sort()
# each element contains in the following order :
# encoder architecture, encoder loss, clustering loss, dataset name, nmi, max nmi
for file_name in stats_files:
arch, enc_loss, clust_loss, dataset_name, dropout, noise = get_elements(file_name)
try:
with open(os.path.join(stat_path, file_name)) as f:
row = f.readline()
# the test's stats are in the second row
if type == 'test':
row = f.readline()
split_stats = row.split(',')
acc = float(split_stats[0])
max_acc = float(split_stats[1])
nmi = float(split_stats[2])
max_nmi = float(split_stats[3])
ari = float(split_stats[4])
max_ari = float(split_stats[5])
gap = max_nmi - nmi
except:
print('error with '+os.path.join(stat_path, file_name))
nmi, max_nmi, gap, ari, acc = -1, -1, 0, -1, -1
stats.append([arch, enc_loss, clust_loss, dataset_name, nmi, max_nmi, gap, ari, acc])
for file_name in error_files:
arch, enc_loss, clust_loss, dataset_name, dropout, noise = get_elements(file_name.split('.')[0])
stats.append([arch, enc_loss, clust_loss, dataset_name, np.nan, np.nan, 0.0, np.nan, np.nan])
stats = np.array(stats)
dfs.append(pd.DataFrame({"encoder_architecture": stats[:, 0],
"encoder_loss": stats[:, 1],
"clustering_loss": stats[:, 2],
"dataset_name": stats[:, 3],
"nmi": stats[:, 4],
"max_nmi": stats[:, 5],
"gap": stats[:, 6],
"ari": stats[:, 7],
"acc": stats[:, 8]}))
concat = pd.concat(dfs)
arch_list = concat.encoder_architecture.unique()
encoder_loss = concat.encoder_loss.unique()
clustering_loss = concat.clustering_loss.unique()
dataset_name = concat.dataset_name.unique()
out_path = root_dir + '/stats_extract/' + str(itr) + '/' + str(seed)
if len(seed_itrs) > 1:
out_path = root_dir + '/stats_extract/' + str(itr) + '/' + str(seed_itrs[0]) + 'to' + str(seed_itrs[-1]) + type
utils.create_directory(root_dir + '/stats_extract/' + str(itr) + '/')
concat["nmi"] = concat["nmi"].astype(float)
concat["max_nmi"] = concat["max_nmi"].astype(float)
concat["gap"] = concat["gap"].astype(float)
concat["acc"] = concat["acc"].astype(float)
concat["ari"] = concat["ari"].astype(float)
concat.to_csv(out_path + '_raw_data.csv', index=False)
concat["model"] = concat["encoder_architecture"] + "_" + concat["encoder_loss"] + "_" + concat["clustering_loss"]
# concat_cleaned = concat.drop(["encoder_architecture", "encoder_loss", "clustering_loss"], axis=1)
# concat_cleaned = concat_cleaned.set_index(["model"])
# df_nmi = df2.drop(["max_nmi"])
df_nmi = pd.pivot_table(concat, values="nmi", index="dataset_name", columns="model", aggfunc=np.nanmean)
df_nmi.to_csv(out_path + '_per_dataset_nmi.csv', index=True)
df_max_nmi = pd.pivot_table(concat, values="max_nmi", index="dataset_name", columns="model", aggfunc=np.nanmean)
df_max_nmi.to_csv(out_path + '_per_dataset_max_nmi.csv', index=True)
df_gap = pd.pivot_table(concat, values="gap", index="dataset_name", columns="model", aggfunc=np.nanmean)
df_gap.to_csv(out_path + '_per_dataset_gap.csv', index=True)
df_gap = pd.pivot_table(concat, values="ari", index="dataset_name", columns="model", aggfunc=np.nanmean)
df_gap.to_csv(out_path + '_per_dataset_ari.csv', index=True)
df_gap = pd.pivot_table(concat, values="acc", index="dataset_name", columns="model", aggfunc=np.nanmean)
df_gap.to_csv(out_path + '_per_dataset_acc.csv', index=True)
if __name__ == '__main__':
parser = argparse.ArgumentParser(
description='Extract stats from logs'
)
parser.add_argument('--itr', type=str, metavar='X', default='50320',
help='iteration index')
parser.add_argument('--seeds_itr', type=str, metavar='X', default='0,1,2,3,4',
help='seeds index, can be either an integer or a comma separated list, '
'example : --seeds_itr 0,1,2,3,4')
parser.add_argument('--root_dir', type=str, metavar='PATH', default='.',
help='path of the root dir where archives and results are stored')
parser.add_argument('--type', type=str, metavar='xxx', default='test',
choices=['train', 'test'],
help='run on either train or test results clustering')
args = parser.parse_args()
itr = args.itr
seed_itrs = args.seeds_itr.split(',')
type = args.type
dfs = []
stat_path = ''
root_dir = args.root_dir
main(itr, seed_itrs, type, stat_path, root_dir)
for manifold_learner in ['UMAP', 'TSNE', 'Isomap', 'LLE', '']:
for method in ['Gmm', 'Kmeans', 'Spectral']:
sub_itr = method+manifold_learner+'/'+itr
print('start '+sub_itr)
main(sub_itr, seed_itrs, type, stat_path, root_dir)