/
run_toy_example.py
112 lines (96 loc) · 4.13 KB
/
run_toy_example.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
import numpy as np
import argparse
from toy_example_data import *
from random_feature_utils import *
def main():
parser = argparse.ArgumentParser()
# Toy example data
parser.add_argument('-n', '--n', type=int, required=True)
parser.add_argument('--toy_example_name', choices=['random_projections','no_projections'], required=True)
# Data generation parameters
parser.add_argument('--p_correlation', type=float)
parser.add_argument('--mean_causal', type=float)
parser.add_argument('--mean_spurious', type=float)
parser.add_argument('--var_causal', type=float)
parser.add_argument('--d_causal', type=int)
parser.add_argument('--d_spurious', type=int)
parser.add_argument('--var_spurious', type=float)
parser.add_argument('--d_noise', type=int)
parser.add_argument('--mean_noise', type=float)
parser.add_argument('--var_noise', type=float)
# Random features
parser.add_argument('-N', '--n_random_features', type=int, action='append')
# Model
parser.add_argument('-m', '--model_type', choices=['ridge','logistic'], required=True)
parser.add_argument('-L', '--Lambda', type=float, default=None)
# Error
parser.add_argument('-e', '--error_type', choices=['zero_one', 'squared'], required=True)
# Outputs
parser.add_argument('-o', '--outfile', required=True)
parser.add_argument('--model_file')
parser.add_argument('-q', '--quiet', action='store_true')
parser.add_argument('-s', '--seed', type=int, default=0)
args = parser.parse_args()
np.random.seed(args.seed)
process_args(args)
if args.random_features:
full_data, n_groups = generate_toy_data(args.data_generation_fn, args.data_args)
erm_error, over_error, under_error = run_random_features_model(full_data=full_data,
n_groups=n_groups,
N=args.n_random_features,
fit_model_fn=args.fit_model_fn,
error_fn=args.error_fn,
model_kwargs=args.model_kwargs,
verbose=(not args.quiet))
else:
erm_error, over_error, under_error = run_no_projection_model(data_generation_fn=args.data_generation_fn,
data_args=args.data_args,
N=args.n_random_features,
fit_model_fn=args.fit_model_fn,
error_fn=args.error_fn,
model_kwargs=args.model_kwargs,
verbose=(not args.quiet),
model_file=args.model_file)
save_error_logs(args.outfile,
[erm_error, over_error, under_error],
['ERM', 'oversample', 'undersample'])
def process_args(args):
# model
if args.model_type=='ridge':
fit_model_fn = fit_ridge_regression
assert args.Lambda
elif args.model_type=='logistic':
fit_model_fn = fit_logistic_regression
model_kwargs = {'Lambda': args.Lambda}
args.fit_model_fn = fit_model_fn
args.model_kwargs = model_kwargs
# data
if args.toy_example_name=='random_projections':
data_generation_fn = generate_toy_data_random_projections
required_args = ['n', 'p_correlation', 'd_causal', 'mean_causal', 'var_causal', 'd_spurious', 'mean_spurious', 'var_spurious']
random_features = True
elif args.toy_example_name=='no_projections':
data_generation_fn = generate_toy_data_no_projections
required_args = ['n', 'p_correlation', 'mean_causal', 'var_causal', 'mean_spurious', 'var_spurious',
'mean_noise', 'var_noise']
random_features = False
assert len(args.n_random_features)>0
if len(args.n_random_features)==1:
args.n_random_features = args.n_random_features[0]
data_args = {}
for argname in required_args:
argval = getattr(args, argname)
assert argval is not None, f'{argname} must be specified'
data_args[argname] = argval
args.data_args = data_args
args.data_generation_fn = data_generation_fn
# error
if args.error_type=='zero_one':
error_fn = zero_one_error
elif args.error_type=='squared':
error_fn = squared_error
args.error_fn = error_fn
# random features?
args.random_features = random_features
if __name__=='__main__':
main()