-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathmain_map.py
190 lines (144 loc) · 5.97 KB
/
main_map.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
#!/usr/bin/env python2
# -*- coding: utf-8 -*-
"""
Created on Tue Mar 20 21:14:04 2018
python main_map.py gamma seed uobs kobs collobs
"""
import sys
import csv
import numpy as np
import matplotlib.pyplot as plt
from pyDOE import lhs
from scipy.interpolate import griddata
#from models_tf import DarcyNet2D_BCs
from test_sdfs_est_lm import *
import tensorflow as tf
tf.set_random_seed(int(sys.argv[-4]))
np.random.seed(int(sys.argv[-4]))
if __name__ == "__main__":
dataset = np.load('./test.npz')
X = dataset['coo']
K = dataset['k']
U = dataset['u']
idx = 11
X_star = X
k_star = K[idx:idx+1,:].T
u_star = U[idx:idx+1,:].T
N_u = int(sys.argv[-3])
N_k = int(sys.argv[-2])
N_f = int(sys.argv[-1])
N = u_star.shape[0]
res = int(N**(1/2))
lb, ub = X.min(0), X.max(0)
#Latin Hypercube sampling
#make obs deterministic for different net initialization seeds and
# increasing as you add more points
np.random.seed(32)
samples=1
idx_us = [np.floor(N*lhs(1, N)).astype(int).flatten()[:N_u] ]*samples
idx_ks = [np.floor(N*lhs(1, N)).astype(int).flatten()[:N_k] ]*samples
# idx_us = [np.floor(N*lhs(1, N)).astype(int).flatten()[:N_u] for _ in
# range(samples)]
# idx_ks = [np.floor(N*lhs(1, N)).astype(int).flatten()[:N_k] for _ in
# range(samples)]
idx_fs = [np.floor(N*lhs(1, N)).astype(int).flatten()[:N_f] ]*samples
# idx_fs = [np.floor(N*lhs(1, N)).astype(int).flatten()[:N_f] for _ in
# range(samples)]
#idx_us = np.apply_along_axis(lambda r : r[0]+32*r[1],
# axis=1,
# arr=np.floor(res*lhs(2)).astype(int))[:N_u]
#reset seed
np.random.seed(int(sys.argv[-4]))
errors_k = []
errors_u = []
for idx_u, idx_k, idx_f in zip(idx_us,idx_ks, idx_fs):
# Training data
# idx_k = np.random.choice(N, N_k)
Y_y = np.log(k_star[idx_k,:]).flatten()
# idx_u = np.random.choice(N, N_u)
Y_u = u_star[idx_u,:].flatten()
L = np.array([1.0, 1.0])
N = np.array([32, 32])
g = Geom(L, N)
g.calculate()
ul = 1.0
ur = 0.0
bc = BC(g)
bc.dirichlet(g, "left", ul)
bc.dirichlet(g, "right", ur)
se = SEKernel(std_dev=1.0, cor_len=0.15, std_dev_noise=1e-4)
CY = se.covar(g.cells.centroids.T, g.cells.centroids.T)
# Create model
prob = DarcyExp(g, bc)
gamma = float(sys.argv[-5])
Lreg = compute_Lreg(g)
# loss = LossVec(idx_u, Y_u, idx_k, Y_y, gamma, spl.inv(spl.cholesky(CY, lower=True)))
loss = LossVec(idx_u, Y_u, idx_k, Y_y, gamma, Lreg)
dasa = DASAExpLM(loss.val, loss.grad_u, loss.grad_Y, prob.solve, prob.residual_sens_u, prob.residual_sens_Y)
Y0 = np.full(g.cells.num, 0.0)
res = spo.leastsq(dasa.obj, Y0, Dfun=dasa.grad)
k_pred = np.exp(res[0]).reshape(-1,1)
# Predict at test points
# Relative L2 error
error_k = np.linalg.norm(k_star - k_pred, 2)/np.linalg.norm(k_star, 2)
# error_u = np.linalg.norm(u_star - u_pred, 2)/np.linalg.norm(u_star, 2)
errors_k.append(error_k)
# errors_u.append(error_u)
# Plot
X_k = X[idx_k,:]
Y_k = k_star[idx_k,:]
nn = 200
x = np.linspace(lb[0], ub[0], nn)
y = np.linspace(lb[1], ub[1], nn)
XX, YY = np.meshgrid(x,y)
K_plot = griddata(X_star, k_pred.flatten(), (XX, YY), method='cubic')
# U_plot = griddata(X_star, u_pred.flatten(), (XX, YY), method='cubic')
K_error = griddata(X_star, np.abs(k_star-k_pred).flatten(), (XX, YY), method='cubic')
# U_error = griddata(X_star, np.abs(u_star-u_pred).flatten(), (XX, YY), method='cubic')
fig = plt.figure(1)
plt.pcolor(XX, YY, K_plot, cmap='viridis')
plt.plot(X_k[:,0], X_k[:,1], 'ro', markersize = 1)
plt.clim(np.min(k_star), np.max(k_star))
plt.colorbar()
plt.xticks(fontsize=14)
plt.yticks(fontsize=14)
plt.xlabel('$x_1$', fontsize=16)
plt.ylabel('$x_2$', fontsize=16)
plt.title('$k(x_1,x_2)$', fontsize=16)
fig.tight_layout()
fig.savefig('./plots/map/map_k_sample_'+str(sys.argv[-5])+'_u_'+sys.argv[-3]+'_k_'+sys.argv[-2]+'_c_'+sys.argv[-1]+'_pred.png')
fig.clf()
fig = plt.figure(3)
plt.pcolor(XX, YY, K_error, cmap='viridis')
plt.colorbar()
plt.xticks(fontsize=14)
plt.yticks(fontsize=14)
plt.xlabel('$x_1$', fontsize=16)
plt.ylabel('$x_2$', fontsize=16)
plt.title('Absolute error', fontsize=16)
fig.tight_layout()
fig.savefig('./plots/map/map_k_sample_'+str(sys.argv[-5])+'_u_'+sys.argv[-3]+'_k_'+sys.argv[-2]+'_c_'+sys.argv[-1]+'_errors.png')
fig.clf()
samples=samples-1
#completely reset tensorflow
tf.reset_default_graph()
with open("./errors/map/map_k_loss_u_"+sys.argv[-3]+"_k_"+sys.argv[-2]+"_c_"+sys.argv[-1]+".csv",
"a") as f:
writer = csv.writer(f, delimiter=',')
writer.writerow(errors_k)
f.close()
# with open("./errors/map_u_loss_u_"+sys.argv[-3]+"_k_"+sys.argv[-2]+"_c_"+sys.argv[-1]+".csv",
# "a") as f:
# writer = csv.writer(f, delimiter=',')
# writer.writerow(errors_u)
# f.close()
#
#
# # filename = sys.argv[-1].replace('/','.').split('.')[-2]
# fig.savefig('./plots/'+sys.argv[-4]+'_u_'+sys.argv[-1]+'_k_'+sys.argv[-2]+'_c_'+sys.argv[-3]+'_test.png')
# with open('./errors/'+sys.argv[-4]+'_u_'+sys.argv[-1]+'_k_'+sys.argv[-2]+'_c_'+sys.argv[-3]+'_k_error_test.txt', 'a') as losses_file:
# print(error_k, file=losses_file)
# losses_file.close()
# with open('./errors/'+sys.argv[-4]+'_u_'+sys.argv[-1]+'_k_'+sys.argv[-2]+'_c_'+sys.argv[-3]+'_u_error_test.txt', 'a') as losses_file:
# print(error_u, file=losses_file)
# losses_file.close()