-
Notifications
You must be signed in to change notification settings - Fork 0
/
real_data_figure_script.py
175 lines (155 loc) · 7.47 KB
/
real_data_figure_script.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
import pickle
from time import time
import numpy as np
import tensorflow as tf
from sklearn.ensemble import RandomForestRegressor
from treatment_frameworks import *
from utility import *
from other_models import *
from funcs import get_ihdp_setup
if __name__ == '__main__':
cur_setup = get_ihdp_setup
train_parts = [0.2, 0.26, 0.32, 0.38, 0.44]
n_splits = 3
alpha = 0.1
batch_size = 256
learning_rate_a = 0.005
learning_rate = 0.01
epochs_num_alpha = 50
epochs_num_ordinary = 50
patience = 5
tasks_alpha = 100000
iters = 1
tasks_num = 100000
n = 100
mlp_coef_val = 200
seed = int(time()) % 2048
ser_name = 'ihdp experiment'
nw_cv_grid = {
'gamma': [10 ** i for i in range(-8, 11)] + [0.5, 5, 50, 100, 200, 500, 700],
}
forest_cv_grid = {
'n_estimators': [10, 50, 100, 300],
'max_depth': [2, 3, 4, 5, 6, 7],
'min_samples_leaf': [1, 0.05, 0.1, 0.2],
}
results = {
'T-NW control mse': [[] for _ in range(len(train_parts))],
'T-NW treat mse': [[] for _ in range(len(train_parts))],
'T-NW CATE mse': [[] for _ in range(len(train_parts))],
'S-NW control mse': [[] for _ in range(len(train_parts))],
'S-NW treat mse': [[] for _ in range(len(train_parts))],
'S-NW CATE mse': [[] for _ in range(len(train_parts))],
'X-NW CATE mse': [[] for _ in range(len(train_parts))],
'T-Learner control mse': [[] for _ in range(len(train_parts))],
'T-Learner treat mse': [[] for _ in range(len(train_parts))],
'T-Learner CATE mse': [[] for _ in range(len(train_parts))],
'S-Learner control mse': [[] for _ in range(len(train_parts))],
'S-Learner treat mse': [[] for _ in range(len(train_parts))],
'S-Learner CATE mse': [[] for _ in range(len(train_parts))],
'X-Learner CATE mse': [[] for _ in range(len(train_parts))],
'Kernel control': [[] for _ in range(len(train_parts))],
'Kernel treat': [[] for _ in range(len(train_parts))],
'Kernel CATE': [[] for _ in range(len(train_parts))],
'time': [],
}
tf.config.set_visible_devices([], 'GPU')
np.seterr(invalid='ignore')
np.random.seed(seed)
try:
for i in range(iters):
seed = np.random.randint(0, 2048)
for j, train_part in enumerate(train_parts):
start_time = time()
setup = cur_setup(seed)
val_part = train_part / (n_splits - 1)
test_part = 1 - train_part - val_part
setup.make_set(val_part, test_part)
train_x, train_y, train_w = setup.get_train_set()
control_x, control_y, treat_x, treat_y = setup.get_cotrol_treat_sets()
test_x, test_control, test_treat, test_cate = setup.get_test_set()
val_set_c, val_labels_c, val_set_t, val_labels_t = setup.get_val_set()
m = train_x.shape[1]
control_size = np.count_nonzero(train_w == 0)
treat_size = np.count_nonzero(train_w == 1)
mlp_coef = tasks_num // control_size
model = get_basic_dynamic_model(
setup, m, n, epochs_num_ordinary, mlp_coef, mlp_coef_val,
learning_rate, seed, batch_size, patience, f_normalize=False)
*test_data_control, cnt_label = make_spec_set(
control_x, test_x, control_y, test_control, control_size, m, 1)
*test_data_treat, trt_label = make_spec_set(treat_x, test_x, treat_y, test_treat, treat_size, m, 1)
cnt_pred = model.predict(test_data_control)
results['Kernel control'][j].append(calc_mse(cnt_pred, cnt_label))
n_t = treat_size // 2
n_c = n_t
alpha_model = get_alpha_dynamic_model(
setup, m, n_c, n_t, epochs_num_alpha, mlp_coef_val,
learning_rate, seed, batch_size, tasks_alpha, alpha, patience, f_normalize=False)
trt_pred_a = alpha_model.predict(test_data_treat)
results['Kernel treat'][j].append(calc_mse(trt_pred_a, trt_label))
results['Kernel CATE'][j].append(
calc_mse(trt_pred_a - cnt_pred, trt_label - cnt_label))
other_models = {
'T-Learner': (make_t_learner, (val_set_c, val_labels_c, val_set_t, val_labels_t, RandomForestRegressor, forest_cv_grid), (train_x, train_y, train_w)),
'S-Learner': (make_s_learner, (val_set_c, val_labels_c, val_set_t, val_labels_t, RandomForestRegressor, forest_cv_grid), (train_x, train_y, train_w)),
'X-Learner': (make_x_learner, (val_set_c, val_labels_c, val_set_t, val_labels_t, RandomForestRegressor, forest_cv_grid), (train_x, train_y, train_w)),
'T-NW': (make_t_learner, (val_set_c, val_labels_c, val_set_t, val_labels_t, KernelRegression, nw_cv_grid), (train_x, train_y, train_w)),
'S-NW': (make_s_learner, (val_set_c, val_labels_c, val_set_t, val_labels_t, KernelRegression, nw_cv_grid), (train_x, train_y, train_w)),
'X-NW': (make_x_learner, (val_set_c, val_labels_c, val_set_t, val_labels_t, KernelRegression, nw_cv_grid), (train_x, train_y, train_w))
}
for key, val in other_models.items():
instance = val[0](*val[1], n_splits, random_state=seed)
instance.fit(*val[2])
if (hasattr(instance, 'predict_control')):
control_pred = instance.predict_control(test_x)
results[f'{key} control mse'][j].append(
calc_mse(control_pred, test_control))
if (hasattr(instance, 'predict_treat')):
treat_pred = instance.predict_treat(test_x)
results[f'{key} treat mse'][j].append(calc_mse(treat_pred, test_treat))
cate_pred = instance.predict(test_x)
results[f'{key} CATE mse'][j].append(calc_mse(cate_pred, test_cate))
print('itetation ', i + 1, '/', iters)
cur_time = time() - start_time
print('time: ', cur_time)
results['time'].append(cur_time)
for key, val in results.items():
if key != 'time':
print(key, ': ', round(val[j][-1], 4))
except KeyboardInterrupt:
pass
# except Exception as e:
# print(e)
# pass
print('time elapsed: ', round(np.sum(results['time']), 0), ' s.')
print('result:')
for key, val in results.items():
if key != 'time':
print(key, ': ', round(np.mean(val), 6))
params = {
'batch_size': batch_size,
'epochs_num_alpha': epochs_num_alpha,
'epochs_num_ordinary': epochs_num_ordinary,
'mlp_coef': mlp_coef,
'patience': patience,
'm': m,
'n': n,
'n_c': n_c,
'n_t': n_t,
'alpha': alpha,
'tasks': tasks_alpha,
'iters': iters,
'control_size': control_size,
'treat_size': treat_size,
'test_size': test_x.shape[0],
'train_parts': train_parts,
'seed': seed,
'learning_rate': learning_rate,
'val_part_c': 0.2,
'val_part_t': 0,
'mlp_coef_val': mlp_coef_val
}
results['params'] = params
with open(f'res_dicts/{ser_name}.pk', 'wb') as file:
pickle.dump(results, file)