-
Notifications
You must be signed in to change notification settings - Fork 1
/
plot_moreau_grad_vs_nof_grads.py
141 lines (107 loc) · 4.62 KB
/
plot_moreau_grad_vs_nof_grads.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
from datetime import datetime
import os
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import numpy as np
from scipy import stats
# from scipy.stats import linregress
def find_dir_with_prefix(prefix, log_dir):
for file in os.listdir(log_dir):
if prefix in file:
return os.path.join(log_dir, file)
raise OSError('dir with prefix \'{}\' not found at: {}'.format(
prefix, log_dir))
if __name__ == '__main__':
if True:
_parent_log_dir = 'results'
subgradvar_log_dir = 'subgrad_var'
subgradvar_prefix = 'd2_m8_SubGradVar_gamma1.1313708498984762_K10000000_x0_4.0_4.0'
subgradvar_seeds = list(range(1, 11))
subgradvar_moreau = []
for seed in subgradvar_seeds:
prefix = 'seed{}_{}'.format(seed, subgradvar_prefix)
parent_log_dir = os.path.join(_parent_log_dir, subgradvar_log_dir)
log_dir = find_dir_with_prefix(prefix, parent_log_dir)
print('seed', seed, ':', log_dir)
npz_file = os.path.join(log_dir, 'output.npz')
npz_file = np.load(npz_file)
sampled_moreau_grad_norm_list = npz_file['sampled_moreau_grad_norm_list']
nof_grads_moreau_list = npz_file['nof_grads_moreau_list']
subgradvar_moreau.append(sampled_moreau_grad_norm_list)
subgradvar_moreau = np.array(subgradvar_moreau)
plt.loglog(
nof_grads_moreau_list, subgradvar_moreau.mean(axis=0),
'b^--', label='Sub-gradient method')
y_error = stats.sem(subgradvar_moreau, axis=0)
plt.errorbar(
(nof_grads_moreau_list),
(subgradvar_moreau).mean(axis=0),
yerr=y_error,
# yerr=np.power(10., y_error),
capsize=3,
fmt = 'b^',
)
plt.legend()
savepath = 'subgradvar_mean.png'
if True:
_parent_log_dir = 'results'
proxfdiag_log_dir = 'prox_fdiag'
proxfdiag_prefix = 'd2_m8_ProxFDIAG_eps'
proxfdiag_seeds = list(range(1, 11))
proxfdiag_epsilons = 10.**np.arange(0, -4, -1)
proxfdiag_moreau_grad_norm = []
proxfdiag_nof_inner_steps = []
for epsilon in proxfdiag_epsilons:
for seed in proxfdiag_seeds:
prefix = 'seed{}_{}{:.3g}'.format(
seed, proxfdiag_prefix, epsilon)
parent_log_dir = os.path.join(_parent_log_dir, proxfdiag_log_dir)
log_dir = find_dir_with_prefix(prefix, parent_log_dir)
print('seed', seed, ':', log_dir)
npz_file = os.path.join(log_dir, 'output.npz')
npz_file = np.load(npz_file)
proxfdiag_moreau_grad_norm.append(npz_file['moreau_grad_norm_list'][-1])
proxfdiag_nof_inner_steps.append(npz_file['nof_grads_list'][-1])
plt.loglog(
proxfdiag_nof_inner_steps, proxfdiag_moreau_grad_norm,
'ro', label='Prox-FDIAG (ours)')
slope, intercept, r_value, p_value, std_err = stats.linregress(
np.log10(proxfdiag_nof_inner_steps), np.log10(proxfdiag_moreau_grad_norm))
inner_steps_range = np.arange(2, 8, 1)
plt.plot(10.**inner_steps_range, 10.**(inner_steps_range*slope+intercept), 'r-')
savepath = 'proxfdiag_mean'
if True:
_parent_log_dir = 'results'
fastprox_log_dir = 'adaptive_prox_fdiag'
fastprox_prefix = 'd2_m8_ProxFDIAG_eps'
fastprox_seeds = list(range(1, 11))
fastprox_moreau_grad_norm = []
fastprox_nof_inner_steps = []
for seed in fastprox_seeds:
prefix = 'seed{}_{}'.format(seed, fastprox_prefix)
parent_log_dir = os.path.join(_parent_log_dir, fastprox_log_dir)
log_dir = find_dir_with_prefix(prefix, parent_log_dir)
print('seed', seed, ':', log_dir)
npz_file = os.path.join(log_dir, 'output.npz')
npz_file = np.load(npz_file)
fastprox_moreau_grad_norm.extend(npz_file['moreau_grad_norm_list'][1:])
fastprox_nof_inner_steps.extend(npz_file['nof_grads_list'][1:])
fastprox_nof_inner_steps = np.array(fastprox_nof_inner_steps).reshape(-1,)
fastprox_moreau_grad_norm = np.array(fastprox_moreau_grad_norm).reshape(-1,)
subset = np.random.choice(
len(fastprox_nof_inner_steps),
int(len(fastprox_nof_inner_steps)))
plt.loglog(
fastprox_nof_inner_steps[subset], fastprox_moreau_grad_norm[subset],
'k.', label='Adaptive Prox-FDIAG (ours)')
slope, intercept, r_value, p_value, std_err = stats.linregress(
np.log10(fastprox_nof_inner_steps), np.log10(fastprox_moreau_grad_norm))
inner_steps_range = np.arange(1, 7, 1)
plt.plot(10.**inner_steps_range, 10.**(inner_steps_range*slope+intercept), 'k-')
savepath = 'proxfdiag_mean'
savepath = os.path.join('results', 'neurips_proxfdiag_vs_subgradvar')
plt.legend()
plt.savefig('{}.png'.format(savepath))
plt.savefig('{}.pdf'.format(savepath), bbox_inches='tight')
plt.close()