-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathhyperparameters.py
267 lines (195 loc) · 18 KB
/
hyperparameters.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
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
def find_params(model_name, dataset_name):
"""SLEARNER"""
params_SLearner_IHDP_a = {'dataset_name': "ihdp_a", 'num': 100, 'lr': 1e-3, 'patience': 40,
'batch_size': 64, 'reg_l2': .01, 'activation': 'linear',
'epochs': 300, 'binary': False, 'n_fc': 9, 'verbose': 0, 'val_split': 0.0,
'kernel_init': 'RandomNormal', 'max_trials': 10}
params_SLearner_IHDP_b = {'dataset_name': "ihdp_b", 'num': 100, 'lr': 1e-3, 'patience': 40,
'batch_size': 64, 'reg_l2': .01, 'activation': 'linear',
'epochs': 300, 'binary': False, 'n_fc': 6, 'verbose': 0, 'val_split': 0.0,
'kernel_init': 'GlorotNormal', 'max_trials': 30}
params_SLearner_ACIC = {'dataset_name': "acic", 'num': 77, 'lr': 1e-3, 'patience': 40,
'batch_size': 256, 'reg_l2': .01, 'activation': 'linear',
'val_split': 0.0, 'epochs': 300, 'binary': False, 'n_fc': 7, 'verbose': 0,
'kernel_init': 'RandomNormal', 'max_trials': 30}
params_SLearner_JOBS = {'dataset_name': "jobs", 'num': 100, 'lr': 1e-3, 'patience': 40,
'batch_size': 128, 'reg_l2': .01, 'activation': 'sigmoid',
'val_split': 0.0, 'epochs': 50, 'binary': True, 'n_fc': 5, 'verbose': 0,
'kernel_init': 'RandomNormal', 'max_trials': 20}
"""TLEARNER"""
params_TLearner_IHDP_a = {'dataset_name': "ihdp_a", 'num': 100, 'lr_0': 1e-2, 'lr_1': 1e-3, 'patience': 40,
'batch_size_0': 64, 'batch_size_1': 64, 'reg_l2': .01, 'activation': 'linear',
'epochs': 300, 'binary': False, 'verbose': 0, 'model_name': 'TLearner',
'kernel_init': 'RandomNormal', 'max_trials': 10}
params_TLearner_IHDP_b = {'dataset_name': "ihdp_b", 'num': 100, 'lr_0': 1e-3, 'lr_1': 1e-3, 'patience': 40,
'batch_size_0': 512, 'batch_size_1': 512, 'reg_l2': .01, 'activation': 'linear',
'epochs': 1200, 'binary': False, 'verbose': 0, 'model_name': 'TLearner',
'kernel_init': 'GlorotNormal', 'max_trials': 10}
params_TLearner_ACIC = {'dataset_name': "acic", 'num': 77, 'lr': 1e-2, 'patience': 40, 'max_trials': 10,
'batch_size_0': 128, 'batch_size_1': 128, 'reg_l2': .01, 'activation': 'linear',
'epochs': 1000, 'binary': False, 'verbose': 0, 'kernel_init': 'GlorotNormal'}
params_TLearner_JOBS = {'dataset_name': "jobs", 'num': 100, 'lr': 1e-2, 'patience': 40,
'max_trials': 20, 'batch_size_0': 256, 'batch_size_1': 256, 'reg_l2': .01,
'activation': 'sigmoid', 'epochs': 30, 'binary': True, 'verbose': 0,
'kernel_init': 'RandomNormal'}
"""RLEARNER"""
params_RLearner_IHDP_a = {'dataset_name': "ihdp_a", 'num': 100, 'lr': 1e-3, 'patience': 40, 'batch_size_g_mu': 256,
'batch_size_r': 32, 'reg_l2': .01, 'activation': 'linear','epochs': 100, 'binary': False,
'val_split': 0.0, 'verbose': 0, 'kernel_init': 'RandomNormal', 'max_trials': 20}
params_RLearner_IHDP_b = {'dataset_name': "ihdp_b", 'num': 100, 'lr': 1e-3, 'patience': 40, 'batch_size_g_mu': 256,
'batch_size_r': 32, 'reg_l2': .01, 'activation': 'linear', 'epochs': 100, 'binary': False,
'val_split': 0.0, 'verbose': 0, 'kernel_init': 'RandomNormal', 'max_trials': 10}
params_RLearner_ACIC = {'dataset_name': "acic", 'num': 10,'lr': 1e-3, 'patience': 40, 'batch_size_g_mu': 256,
'batch_size_r': 128, 'reg_l2': .01, 'activation': 'linear', 'epochs': 300, 'binary': False,
'val_split': 0.0, 'verbose': 0, 'kernel_init': 'GlorotNormal', 'max_trials': 10}
params_RLearner_JOBS = {'dataset_name': "jobs", 'num': 100, 'lr': 1e-4, 'patience': 40, 'batch_size_g_mu': 512,
'batch_size': 32, 'reg_l2': .001, 'activation': 'sigmoid', 'epochs': 30, 'binary': True,
'val_split': 0.0, 'verbose': 0, 'kernel_init': 'GlorotNormal', 'max_trials': 10}
"""XLEARNER"""
params_XLearner_IHDP_a = {'dataset_name': "ihdp_a", 'num': 100, 'lr': 1e-3, 'patience': 40, 'batch_size_g_mu': 256,
'batch_size': 673, 'reg_l2': .01, 'activation': 'linear', 'epochs': 300, 'binary': False,
'kernel_init': 'RandomNormal','val_split': 0.0, 'verbose': 0}
params_XLearner_IHDP_b = {'dataset_name': "ihdp_b", 'num': 100, 'lr': 1e-2, 'patience': 40, 'batch_size_g_mu': 256,
'batch_size': 256, 'reg_l2': .01, 'activation': 'linear', 'epochs': 200, 'binary': False,
'kernel_init': 'RandomNormal', 'hidden_d': 128, 'verbose': 0}
params_XLearner_ACIC = {'dataset_name': "acic", 'num': 100, 'lr': 1e-3, 'patience': 40, 'batch_size_g_mu': 256,
'batch_size': 673, 'reg_l2': .01, 'activation': None, 'epochs': 1000, 'binary': False,
'verbose': 0, 'kernel_init': 'RandomNormal', 'val_split': 0.0}
params_XLearner_JOBS = {'dataset_name': "jobs", 'num': 100, 'lr': 1e-3, 'patience': 40, 'batch_size_g_mu': 256,
'batch_size': 673, 'reg_l2': .01, 'activation': 'sigmoid', 'epochs': 1000, 'binary': True,
'verbose': 0, 'kernel_init': 'GlorotNormal', 'val_split': 0.0}
"""TARNET"""
params_TARnet_IHDP_a = {'dataset_name': "ihdp_a", 'num': 100, 'lr': 1e-4, 'patience': 40, 'batch_size': 64,
'reg_l2': .01, 'activation': 'linear', 'epochs': 300, 'binary': False, 'verbose': 0,
'val_split': 0.0, 'kernel_init': 'RandomNormal', 'max_trials': 10}
params_TARnet_IHDP_b = {'dataset_name': "ihdp_b", 'num': 100, 'lr': 1e-4, 'patience': 5, 'batch_size': 32,
'reg_l2': .01, 'activation': 'linear', 'epochs': 300, 'binary': False, 'verbose': 0,
'val_split': 0.0, 'kernel_init': 'GlorotNormal', 'max_trials': 10}
params_TARnet_ACIC = {'dataset_name': "acic", 'num': 77, 'lr': 1e-3, 'patience': 40, 'batch_size': 256,
'reg_l2': .01, 'activation': 'linear', 'epochs': 300, 'binary': False, 'verbose': 0,
'val_split': 0.0, 'kernel_init': 'RandomNormal', 'max_trials': 10}
params_TARnet_JOBS = {'dataset_name': "jobs", 'num': 100, 'lr': 1e-2, 'patience': 40, 'batch_size': 256,
'reg_l2': .01, 'activation': 'sigmoid', 'epochs': 30, 'binary': True,
'val_split': 0.0, 'verbose': 0, 'kernel_init': 'GlorotNormal', 'max_trials': 20}
"""DRAGONNET"""
params_DragonNet_IHDP_a = {'dataset_name': "ihdp_a", 'num': 100, 'lr': 1e-4, 'patience': 40, 'batch_size': 64,
'reg_l2': .01, 'activation': 'linear', 'epochs': 300, 'binary': False,
'verbose': 0, 'kernel_init': 'RandomNormal', 'val_split': 0.0, 'max_trials': 10}
params_DragonNet_IHDP_b = {'dataset_name': "ihdp_b", 'num': 100, 'lr': 1e-4, 'patience': 5, 'batch_size': 32,
'reg_l2': .01, 'activation': 'linear', 'epochs': 300, 'binary': False, 'verbose': 0,
'kernel_init': 'GlorotNormal', 'val_split': 0.0, 'max_trials': 10}
params_DragonNet_ACIC = {'dataset_name': "acic", 'num': 77, 'lr': 1e-4, 'patience': 40, 'batch_size': 256,
'reg_l2': .01, 'activation': 'linear', 'epochs': 300, 'binary': False, 'val_split': 0.0,
'verbose': 0, 'kernel_init': 'RandomNormal', 'max_trials': 10}
params_DragonNet_JOBS = {'dataset_name': "jobs", 'num': 100, 'lr': 1e-3, 'patience': 40, 'batch_size': 512,
'reg_l2': .01, 'activation': 'sigmoid', 'epochs': 30, 'binary': True, 'val_split': 0.0,
'verbose': 0, 'kernel_init': 'RandomNormal', 'max_trials': 20}
"""CEVAE"""
params_CEVAE_IHDP_a = {'dataset_name': "ihdp_a", 'num': 100, 'num_bin': 19, 'num_cont': 6, 'lr': 1e-3,
'patience': 40, 'batch_size': 64, 'reg_l2': .01, 'activation': 'linear',
'latent_dim': 20, 'epochs': 300, 'binary': False,'val_split': 0.0, 'verbose': 0,
'kernel_init': 'RandomNormal'}
params_CEVAE_IHDP_b = {'dataset_name': "ihdp_b", 'num': 100, 'num_bin': 19, 'num_cont': 6, 'lr': 1e-3,
'patience': 40, 'batch_size': 64, 'reg_l2': .01, 'activation': 'linear',
'latent_dim': 20, 'epochs': 300, 'binary': False, 'val_split': 0.0, 'verbose': 0,
'kernel_init': 'GlorotNormal'}
params_CEVAE_ACIC = {'dataset_name': "acic", 'num': 77, 'num_bin': 55, 'num_cont': 0, 'lr': 1e-3, 'patience': 40,
'batch_size': 400, 'reg_l2': .01, 'activation': 'linear', 'latent_dim': 20,
'epochs': 1000, 'binary': False, 'val_split': 0.0, 'verbose': 0,
'kernel_init': 'RandomNormal'}
params_CEVAE_JOBS = {'dataset_name': "jobs", 'num': 100, 'lr': 1e-3, 'num_bin': 0, 'num_cont': 17, 'patience': 40,
'batch_size': 1024, 'reg_l2': .01, 'activation': 'sigmoid', 'latent_dim': 20,
'epochs': 50, 'binary': True, 'val_split': 0.0, 'verbose': 0,
'kernel_init': 'GlorotNormal'}
"""TEDVAE"""
params_TEDVAE_IHDP_a = {'dataset_name': "ihdp_a", 'num': 100, 'lr': 1e-4, 'patience': 40, 'batch_size': 1024,
'reg_l2': .01, 'activation': 'linear', 'latent_dim_z': 15,
'num_bin': 19,'num_cont': 6, 'latent_dim_zt': 15,
'latent_dim_zy': 5, 'epochs': 400, 'binary': False,
'verbose': 0, 'kernel_init': 'RandomNormal'}
params_TEDVAE_IHDP_b = {'dataset_name': "ihdp_b", 'num': 100, 'lr': 1e-3, 'patience': 40, 'batch_size': 1024,
'reg_l2': .01, 'activation': 'linear', 'latent_dim_z': 15, 'num_bin': 19,
'num_cont': 6, 'latent_dim_zt': 15, 'latent_dim_zy': 5, 'epochs': 400, 'binary': False,
'verbose': 0, 'val_split': 0.0, 'kernel_init': 'GlorotNormal'}
params_TEDVAE_ACIC = {'dataset_name': "acic", 'num': 77, 'lr': 1e-3, 'patience': 25, 'batch_size': 32,
'reg_l2': .01, 'activation': 'linear', 'latent_dim_z': 15, 'num_bin': 55,
'num_cont': 0, 'latent_dim_zt': 15, 'latent_dim_zy': 5, 'epochs': 300, 'binary': False,
'val_split': 0.0, 'verbose': 0, 'kernel_init': 'RandomNormal'}
params_TEDVAE_JOBS = {'dataset_name': "jobs", 'num': 100, 'lr': 1e-3, 'patience': 40, 'batch_size': 256,
'reg_l2': .01, 'activation': 'sigmoid', 'latent_dim_z': 15, 'num_bin': 0,
'num_cont': 17, 'latent_dim_zt': 15, 'latent_dim_zy': 5, 'epochs': 30, 'binary': True,
'val_split': 0.0, 'verbose': 0, 'kernel_init': 'GlorotNormal'}
"""CFRNET"""
params_CFRNet_IHDP_a = {'dataset_name': "ihdp_a", 'num': 100, 'lr': 1e-4, 'patience': 40, 'tuner_batch_size': 256,
'batch_size': 1024, 'hidden_phi': 200, 'reg_l2': .01, 'activation': 'linear', 'epochs': 300,
'binary': False, 'verbose': 0, 'kernel_init': 'RandomNormal'}
params_CFRNet_IHDP_b = {'dataset_name': "ihdp_b", 'num': 100, 'lr': 1e-3, 'patience': 5, 'hidden_phi': 200,
'tuner_batch_size': 256, 'batch_size': 1024, 'reg_l2': .01, 'activation': 'linear',
'epochs': 300, 'binary': False, 'verbose': 0, 'kernel_init': 'GlorotNormal'}
params_CFRNet_ACIC = {'dataset_name': "acic", 'num': 77, 'lr': 1e-3, 'patience': 40, 'tuner_batch_size': 256,
'batch_size': 512, 'reg_l2': .01, 'activation': 'linear', 'epochs': 500, 'binary': False,
'verbose': 0, 'kernel_init': 'GlorotNormal'}
params_CFRNet_JOBS = {'dataset_name': "jobs", 'num': 100, 'lr': 1e-2, 'patience': 40, 'tuner_batch_size': 512,
'hidden_phi': 200, 'batch_size': 1024, 'reg_l2': .01, 'activation': 'sigmoid', 'epochs': 50,
'binary': True, 'verbose': 0, 'kernel_init': 'RandomNormal'}
"""GANITE"""
params_GANITE_IHDP_a = {'dataset_name': "ihdp_a", 'num': 100, 'lr': 1e-3, 'patience': 40,
'batch_size_g': 64, 'batch_size_i': 128, 'reg_l2': .01, 'activation': 'linear',
'binary': False, 'epochs_g': 1000, 'verbose': 0, 'epochs_i': 500, 'val_split': 0.0,
'kernel_init': 'RandomNormal'}
params_GANITE_IHDP_b = {'dataset_name': "ihdp_b", 'num': 100, 'lr': 1e-3, 'patience': 40, 'batch_size_g': 64,
'batch_size_i': 256, 'reg_l2': .01, 'activation': 'linear', 'binary': False,
'epochs_g': 1000, 'verbose': 0, 'epochs_i': 500, 'val_split': 0.0,
'kernel_init': 'GlorotNormal'}
params_GANITE_ACIC = {'dataset_name': "acic", 'num': 77, 'lr': 1e-4, 'patience': 40,'batch_size_g': 128,
'batch_size_i': 256, 'reg_l2': .01, 'activation': 'linear', 'binary': False, 'epochs_g': 1000,
'verbose': 0, 'epochs_i': 500, 'val_split': 0.0, 'kernel_init': 'RandomNormal'}
params_GANITE_JOBS = {'dataset_name': "jobs", 'num': 100, 'lr': 1e-3, 'patience': 40, 'batch_size_g': 64,
'batch_size_i': 256, 'reg_l2': .01, 'activation': 'linear', 'binary': True, 'epochs_g': 1000,
'verbose': 0, 'epochs_i': 500, 'val_split': 0.0, 'kernel_init': 'RandomNormal'}
"""DKLITE"""
params_DKLITE_IHDP_a = {'dataset_name': "ihdp_a", 'num': 100, 'lr': 1e-3, 'patience': 40, 'batch_size': 1024,
'max_trials': 10, 'tuner_batch_size': 32, 'reg_l2': .01, 'activation': 'linear',
'dim_z': 80, 'epochs': 150, 'binary': False, 'reg_var': 1.0, 'reg_rec': 0.7,
'verbose': 0, 'kernel_init': 'RandomNormal', 'x_size': 25}
params_DKLITE_IHDP_b = {'dataset_name': "ihdp_a", 'num': 100, 'lr': 1e-3, 'patience': 40, 'batch_size': 1024,
'max_trials': 15, 'tuner_batch_size': 1024, 'reg_l2': .01, 'activation': 'linear',
'dim_z': 80, 'epochs': 300, 'binary': False, 'reg_var': 1.0, 'reg_rec': 0.7,
'verbose': 0, 'kernel_init': 'GlorotNormal', 'x_size': 25}
params_DKLITE_ACIC = {'dataset_name': "acic", 'num': 77, 'lr': 1e-3, 'patience': 40, 'batch_size': 32,
'max_trials': 20, 'tuner_batch_size': 32, 'reg_l2': .01, 'activation': 'linear',
'dim_z': 50, 'epochs': 300, 'binary': False, 'reg_var': 1.0, 'reg_rec': 0.7,
'verbose': 0, 'kernel_init': 'RandomNormal', 'x_size': 55}
params_DKLITE_JOBS = {'dataset_name': "jobs", 'num': 100, 'lr': 1e-4, 'patience': 40, 'batch_size': 512,
'max_trials': 10, 'tuner_batch_size': 512, 'reg_l2': .01, 'activation': 'sigmoid',
'dim_z': 80, 'epochs': 30, 'binary': True, 'reg_var': 1.0, 'reg_rec': 0.7,
'verbose': 0, 'kernel_init': 'GlorotNormal', 'x_size': 17}
"""-------------------------------------------------------------"""
params_TARnet = {'ihdp_a': params_TARnet_IHDP_a, 'ihdp_b': params_TARnet_IHDP_b, 'acic': params_TARnet_ACIC,
'jobs': params_TARnet_JOBS}
params_CEVAE = {'ihdp_a': params_CEVAE_IHDP_a, 'ihdp_b': params_CEVAE_IHDP_b, 'acic': params_CEVAE_ACIC,
'jobs': params_CEVAE_JOBS}
params_TEDVAE = {'ihdp_a': params_TEDVAE_IHDP_a, 'ihdp_b': params_TEDVAE_IHDP_b, 'acic': params_TEDVAE_ACIC,
'jobs': params_TEDVAE_JOBS}
params_DKLITE = {'ihdp_a': params_DKLITE_IHDP_a, 'ihdp_b': params_DKLITE_IHDP_b, 'acic': params_DKLITE_ACIC,
'jobs': params_DKLITE_JOBS}
params_GANITE = {'ihdp_a': params_GANITE_IHDP_a, 'ihdp_b': params_GANITE_IHDP_b, 'acic': params_GANITE_ACIC,
'jobs': params_GANITE_JOBS}
params_DragonNet = {'ihdp_a': params_DragonNet_IHDP_a, 'ihdp_b': params_DragonNet_IHDP_b,
'acic': params_DragonNet_ACIC, 'jobs': params_DragonNet_JOBS}
params_TLearner = {'ihdp_a': params_TLearner_IHDP_a, 'ihdp_b': params_TLearner_IHDP_b, 'acic': params_TLearner_ACIC,
'jobs': params_TLearner_JOBS}
params_SLearner = {'ihdp_a': params_SLearner_IHDP_a, 'ihdp_b': params_SLearner_IHDP_b, 'acic': params_SLearner_ACIC,
'jobs': params_SLearner_JOBS}
params_RLearner = {'ihdp_a': params_RLearner_IHDP_a, 'ihdp_b': params_RLearner_IHDP_b, 'acic': params_RLearner_ACIC,
'jobs': params_RLearner_JOBS}
params_XLearner = {'ihdp_a': params_XLearner_IHDP_a, 'ihdp_b': params_XLearner_IHDP_b, 'acic': params_XLearner_ACIC,
'jobs': params_XLearner_JOBS}
params_CFRNet = {'ihdp_a': params_CFRNet_IHDP_a, 'ihdp_b': params_CFRNet_IHDP_b, 'acic': params_CFRNet_ACIC,
'jobs': params_CFRNet_JOBS}
"""-------------------------------------------------------------"""
params = {'TARnet': params_TARnet, 'CEVAE': params_CEVAE, 'TEDVAE': params_TEDVAE, 'DKLITE': params_DKLITE,
'DragonNet': params_DragonNet, 'TLearner': params_TLearner, 'SLearner': params_SLearner,
'RLearner': params_RLearner, 'XLearner': params_XLearner, 'CFRNet': params_CFRNet,
'GANITE': params_GANITE}
return params[model_name][dataset_name]