-
Notifications
You must be signed in to change notification settings - Fork 4
/
Oracle_Training.py
219 lines (173 loc) · 8.03 KB
/
Oracle_Training.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
import numpy as np
from SparseARD import *
tol=1e-5
verbose_ard = False
def Train_ARDr(Theta, y, xi_true, ARD_results=None, lams = None, eta = None, verbose = False):
if lams is None: lams = [np.linalg.norm(y)**2*l for l in [0,1e-3,1e-2,1e-1,1e0,1e1,1e2]]
if eta is None: eta = 0.1
if ARD_results is None:
gamma_ARD,sigma2_ARD,mu_xi_ARD,_ = SBL([Theta,y], \
sigma2=1, \
estimate_sigma=True, \
maxit=500, \
verbose=verbose_ard, \
tol=tol)
L0_err_ARD = sparsity_err(mu_xi_ARD, xi_true)[0]
else:
gamma_ARD,sigma2_ARD,mu_xi_ARD = ARD_results
L0_err_ARD = sparsity_err(mu_xi_ARD, xi_true)[0]
Gamma_ARDr = [gamma_ARD]
Sigma2_ARDr = [sigma2_ARD]
Mu_xi_ARDr = [mu_xi_ARD]
L0_err = [L0_err_ARD]
for lam in lams[1:]:
if verbose: print('Training with lambda='+str(lam), end = '. ')
gamma,sigma2,mu_xi,_ = SBL([Theta,y], \
gamma=gamma_ARD, \
lam=lam, \
eta=eta, \
sigma2=sigma2_ARD, \
maxit = 500, \
tol = tol, \
verbose = verbose_ard, \
estimate_sigma=True, \
regularize = True)
Gamma_ARDr.append(gamma)
Sigma2_ARDr.append(sigma2)
Mu_xi_ARDr.append(mu_xi)
L0_err.append(sparsity_err(mu_xi, xi_true)[0])
if verbose: print('L0_err='+str(L0_err[-1]))
q = np.argmin(L0_err)
gamma_ARDr = Gamma_ARDr[q]
sigma2_ARDr = Sigma2_ARDr[q]
mu_xi_ARDr = Mu_xi_ARDr[q]
return gamma_ARDr, sigma2_ARDr, mu_xi_ARDr
def Train_ARDvi(Theta, y, xi_true, ARD_results=None, alphas = None, verbose = False):
if alphas is None: alphas = [1,2,4,8,10,16,32,64,128]
if ARD_results is None:
gamma_ARD,sigma2_ARD,mu_xi_ARD,_ = SBL([Theta,y], \
sigma2=1, \
estimate_sigma=True, \
maxit=500, \
verbose=verbose_ard, \
tol=tol)
L0_err_ARD = sparsity_err(mu_xi_ARD, xi_true)[0]
else:
gamma_ARD,sigma2_ARD,mu_xi_ARD = ARD_results
L0_err_ARD = sparsity_err(mu_xi_ARD, xi_true)[0]
Gamma_ARDvi = [gamma_ARD]
Sigma2_ARDvi = [sigma2_ARD]
Mu_xi_ARDvi = [mu_xi_ARD]
L0_err = [L0_err_ARD]
for alpha in alphas[1:]:
if verbose: print('Training with alpha='+str(alpha), end = '. ')
gamma, sigma2, mu_xi, _ = SBL([Theta,y], \
sigma2=1, \
estimate_sigma=True, \
maxit=500, \
verbose=verbose_ard, \
tol=tol, \
alpha = alpha)
Gamma_ARDvi.append(gamma)
Sigma2_ARDvi.append(sigma2)
Mu_xi_ARDvi.append(mu_xi)
L0_err.append(sparsity_err(mu_xi, xi_true)[0])
if verbose: print('L0_err='+str(L0_err[-1]))
q = np.argmin(L0_err)
gamma_ARDvi = Gamma_ARDvi[q]
sigma2_ARDvi = Sigma2_ARDvi[q]
mu_xi_ARDvi = Mu_xi_ARDvi[q]
return gamma_ARDvi, sigma2_ARDvi, mu_xi_ARDvi
def Train_M_STSBL(Theta, y, xi_true, ARD_results=None, taus = None, verbose = False):
if taus is None: taus = [0,1e-5,1e-4,1e-3,1e-2,1e-1,1]
if ARD_results is None:
gamma_ARD,sigma2_ARD,mu_xi_ARD,_ = SBL([Theta,y], \
sigma2=1, \
estimate_sigma=True, \
maxit=500, \
verbose=verbose_ard, \
tol=tol)
L0_err_ARD = sparsity_err(mu_xi_ARD, xi_true)[0]
else:
gamma_ARD,sigma2_ARD,mu_xi_ARD = ARD_results
L0_err_ARD = sparsity_err(mu_xi_ARD, xi_true)[0]
Gamma_STSBL = [gamma_ARD]
Sigma2_STSBL = [sigma2_ARD]
Mu_xi_STSBL = [mu_xi_ARD]
L0_err = [L0_err_ARD]
for tau in taus[1:]:
if verbose: print('Training with tau='+str(tau), end = '. ')
mu_xi, gamma, sigma2 = M_STSBL(Theta, y, tau, tol=tol, verbose=verbose_ard, \
warm_start = [np.copy(gamma_ARD), np.copy(sigma2_ARD), np.copy(mu_xi_ARD)])
Gamma_STSBL.append(gamma)
Sigma2_STSBL.append(sigma2)
Mu_xi_STSBL.append(mu_xi)
L0_err.append(sparsity_err(mu_xi, xi_true)[0])
if verbose: print('L0_err='+str(L0_err[-1]))
q = np.argmin(L0_err)
gamma_STSBL = Gamma_STSBL[q]
sigma2_STSBL = Sigma2_STSBL[q]
mu_xi_STSBL = Mu_xi_STSBL[q]
return gamma_STSBL, sigma2_STSBL, mu_xi_STSBL
def Train_L_STSBL(Theta, y, xi_true, ARD_results=None, taus = None, verbose = False):
if taus is None: taus = [np.inf,1e-4,1e-3,1e-2,1e-1,1,2,5,10]
if ARD_results is None:
gamma_ARD,sigma2_ARD,mu_xi_ARD,_ = SBL([Theta,y], \
sigma2=1, \
estimate_sigma=True, \
maxit=500, \
verbose=verbose_ard, \
tol=tol)
L0_err_ARD = sparsity_err(mu_xi_ARD, xi_true)[0]
else:
gamma_ARD,sigma2_ARD,mu_xi_ARD = ARD_results
L0_err_ARD = sparsity_err(mu_xi_ARD, xi_true)[0]
Gamma_L_STSBL = [gamma_ARD]
Sigma2_L_STSBL = [sigma2_ARD]
Mu_xi_L_STSBL = [mu_xi_ARD]
L0_err = [L0_err_ARD]
for tau in taus[1:]:
if verbose: print('Training with tau='+str(tau), end = '. ')
mu_xi, gamma, sigma2 = L_STSBL(Theta, y, tau, tol=tol, verbose=verbose_ard, \
warm_start = [np.copy(gamma_ARD), np.copy(sigma2_ARD), np.copy(mu_xi_ARD)])
Gamma_L_STSBL.append(gamma)
Sigma2_L_STSBL.append(sigma2)
Mu_xi_L_STSBL.append(mu_xi)
L0_err.append(sparsity_err(mu_xi, xi_true)[0])
if verbose: print('L0_err='+str(L0_err[-1]))
q = np.argmin(L0_err)
gamma_L_STSBL = Gamma_L_STSBL[q]
sigma2_L_STSBL = Sigma2_L_STSBL[q]
mu_xi_L_STSBL = Mu_xi_L_STSBL[q]
return gamma_L_STSBL, sigma2_L_STSBL, mu_xi_L_STSBL
def Train_MAP_STSBL(Theta, y, xi_true, ARD_results=None, taus = None, verbose = False):
if taus is None: taus = [0,1e-2,1e-1,1,1e1,1e2,1e3]
if ARD_results is None:
gamma_ARD,sigma2_ARD,mu_xi_ARD,_ = SBL([Theta,y], \
sigma2=1, \
estimate_sigma=True, \
maxit=500, \
verbose=verbose_ard, \
tol=tol)
L0_err_ARD = sparsity_err(mu_xi_ARD, xi_true)[0]
else:
gamma_ARD,sigma2_ARD,mu_xi_ARD = ARD_results
L0_err_ARD = sparsity_err(mu_xi_ARD, xi_true)[0]
Gamma_d_STSBL = [gamma_ARD]
Sigma2_d_STSBL = [sigma2_ARD]
Mu_xi_d_STSBL = [mu_xi_ARD]
L0_err = [L0_err_ARD]
for tau in taus[1:]:
if verbose: print('Training with tau='+str(tau), end = '. ')
mu_xi, gamma, sigma2 = MAP_STSBL(Theta, y, tau, 1, tol=tol, verbose=verbose_ard, \
warm_start = [np.copy(gamma_ARD), np.copy(sigma2_ARD), np.copy(mu_xi_ARD)])
Gamma_d_STSBL.append(gamma)
Sigma2_d_STSBL.append(sigma2)
Mu_xi_d_STSBL.append(mu_xi)
L0_err.append(sparsity_err(mu_xi, xi_true)[0])
if verbose: print('L0_err='+str(L0_err[-1]))
q = np.argmin(L0_err)
gamma_d_STSBL = Gamma_d_STSBL[q]
sigma2_d_STSBL = Sigma2_d_STSBL[q]
mu_xi_d_STSBL = Mu_xi_d_STSBL[q]
return gamma_d_STSBL, sigma2_d_STSBL, mu_xi_d_STSBL