-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmain_real.py
207 lines (137 loc) · 6.77 KB
/
main_real.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
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
#!/usr/bin/env python
# -*- coding: utf-8 -*-
import logging
from pytz import timezone
from datetime import datetime
import numpy as np
import torch
from data_loader.real_dataset import RealDataset
from models.Hierarchical_Causal_Clustering import Hierarchical_Causal_Clustering
from trainers.trainer import Trainer
from helpers.config_utils import save_yaml_config, get_args
from helpers.log_helper import LogHelper
from helpers.torch_utils import set_seed
from helpers.dir_utils import create_dir
from helpers.analyze_utils import plot_timeseries, plot_recovered_graph
from tslearn.clustering import TimeSeriesKMeans
from sklearn.metrics import adjusted_rand_score
def kmeans_dtw(X,num_groups,cluster_labels):
_logger = logging.getLogger(__name__)
km = TimeSeriesKMeans(n_clusters=num_groups, metric="dtw")
km_dtw_labels = km.fit_predict(X=X)
ari=adjusted_rand_score(cluster_labels, km_dtw_labels)
return ari,km_dtw_labels
def kmeans_euclidean(X,num_groups,cluster_labels):
_logger = logging.getLogger(__name__)
km = TimeSeriesKMeans(n_clusters=num_groups, metric="euclidean")
km_euclidean_labels = km.fit_predict(X=X)
ari=adjusted_rand_score(cluster_labels, km_euclidean_labels)
return ari,km_euclidean_labels
def DBSCAN_dtw(X, cluster_labels):
_logger = logging.getLogger(__name__)
from tslearn.metrics import cdist_dtw
X_distance = cdist_dtw(X)
from sklearn.cluster import DBSCAN
db_dtw = DBSCAN(metric='precomputed',eps=X_distance.mean() )
DBSCAN_dtw_labels = db_dtw.fit_predict(X=X_distance)
ari=adjusted_rand_score(cluster_labels, DBSCAN_dtw_labels)
return ari,DBSCAN_dtw_labels
def OPTICS_dtw(X, cluster_labels):
_logger = logging.getLogger(__name__)
from tslearn.metrics import cdist_dtw
X_distance = cdist_dtw(X)
from sklearn.cluster import OPTICS
op_dtw = OPTICS(metric='precomputed')
OPTICS_dtw_labels = op_dtw.fit_predict(X=X_distance)
ari=adjusted_rand_score(cluster_labels, OPTICS_dtw_labels)
return ari,OPTICS_dtw_labels
def baseline_cell(output_dir):
np.set_printoptions(precision=3)
_logger = logging.getLogger(__name__)
# Get arguments parsed
args = get_args()
args.num_samples = 30
args.num_subjects_per_group = 25
args.num_groups = 2
args.num_variables = 11
# Reproducibility
set_seed(args.seed)
dataset = RealDataset().cell_dataset
# Shuffle the order of subjects
data = np.concatenate(dataset)
permutation = np.random.permutation(data.shape[0])
shuffled_data = data[permutation]
shuffled_cluster = np.concatenate([ [i]*args.num_subjects_per_group for i in range(args.num_groups) ]) [permutation]
# Baseline methods
ari,_ = kmeans_dtw(shuffled_data,args.num_groups,shuffled_cluster)
_logger.info('Using k-means(DTW) and Adjusted Rand index，ARI: {}'.format(ari))
ari,_ = kmeans_euclidean(shuffled_data,args.num_groups,shuffled_cluster)
_logger.info('Using k-means(euclidean) and Adjusted Rand index，ARI: {}'.format(ari))
ari,_ = DBSCAN_dtw(shuffled_data,shuffled_cluster)
_logger.info('Using DBSCAN(DTW) and Adjusted Rand index，ARI: {}'.format(ari))
ari,_ = OPTICS_dtw(shuffled_data,shuffled_cluster)
_logger.info('Using OPTICS(DTW) and Adjusted Rand index，ARI: {}'.format(ari))
def lead_cell(output_dir):
np.set_printoptions(precision=3)
# Get arguments parsed
args = get_args()
args.num_samples = 30
args.num_subjects_per_group = 25
args.num_groups = 2
args.num_variables = 11
args.max_lag = 0
args.num_iterations_clustering = 1000
args.num_iterations_structurelearning = 2000
args.num_total_iterations = 2
args.seed=2022
_logger = logging.getLogger(__name__)
# Save the configuration for logging purpose
save_yaml_config(args, path='{}/config_cell.yaml'.format(output_dir))
# Reproducibility
set_seed(args.seed)
dataset = RealDataset().cell_dataset
# Look at data
_logger.info('The shape of data (num_groups ,num_subjects, Ts, num_variables): {}, {}'.format(len(dataset),dataset[0].shape))
# for q in range(args.num_groups):
# plot_timeseries(dataset[q],'Group {}'.format(q),display_mode=False,save_name=output_dir+'/group{}_timeseries_cell.png'.format(q+1))
# Shuffle the order of subjects
data = np.concatenate(dataset)
permutation = np.random.permutation(data.shape[0])
shuffled_data = data[permutation]
shuffled_cluster = np.concatenate([ [i]*args.num_subjects_per_group for i in range(args.num_groups) ]) [permutation]
_logger.info('The groudtruth clusters: {}'.format(shuffled_cluster))
# Init model
model = Hierarchical_Causal_Clustering(args.num_samples, args.num_variables, args.max_lag, args.device, args.prior_mu, args.prior_sigma, args.prior_nu, args.prior_omega)
trainer = Trainer(args.learning_rate, args.num_iterations_clustering, args.num_iterations_structurelearning, args.num_output, args.num_MC_sample, args.num_total_iterations)
input_X = torch.tensor(shuffled_data,dtype=torch.float32,device=args.device)
trainer.train_model(model=model, X = input_X, output_dir=output_dir)
# Save result
trainer.log_and_save_intermediate_outputs(model)
_logger.info('Finished training model')
_logger.info('The groudtruth clusters: {}'.format(shuffled_cluster))
# Calculate performance
estimated_cluster = [-1]*(args.num_groups * args.num_subjects_per_group)
for c in model.cluster:
for subject in c:
estimated_cluster[subject] = model.cluster.index(c)
_logger.info('The estimated clusters: {}'.format(estimated_cluster))
from sklearn.metrics import adjusted_rand_score
ari=adjusted_rand_score(shuffled_cluster, estimated_cluster)
_logger.info('Adjusted Rand index，ARI: {}'.format(ari))
n_cluster = len(model.cluster)
for c in range(n_cluster):
parameters = model.causal_structures[c]
estimate_B = np.abs(parameters[0][0].numpy())
threshold = sorted(estimate_B.reshape(-1))[-10] # Top N edges
estimate_B[estimate_B<threshold]=0
estimate_graph = estimate_B.reshape(1,args.num_variables,args.num_variables)
group_dag = None
plot_recovered_graph(estimate_graph,group_dag,title='estimate_graph_{} in cellular dataset'.format(c),display_mode=False,save_name=output_dir+'/estimate_graph_{} in cellular dataset.png'.format(c))
if __name__ == '__main__':
output_dir = 'output/{}'.format(datetime.now(timezone('Asia/Shanghai')).strftime('%Y-%m-%d_%H-%M-%S-%f')[:-3])
create_dir(output_dir)
# Setup for logging
LogHelper.setup(log_path='{}/training.log'.format(output_dir),
level_str='INFO')
lead_cell(output_dir)
baseline_cell(output_dir)