-
Notifications
You must be signed in to change notification settings - Fork 3
/
main_mnist.py
132 lines (106 loc) · 3.47 KB
/
main_mnist.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
#pylint: skip-file
import time
import sys
import numpy as np
import theano
import theano.tensor as T
from AVAE import *
import data
import matplotlib.pyplot as plt
use_gpu(0)
lr = 0.001
drop_rate = 0.
batch_size = 128
hidden_size = 400
latent_size = 2
iter_d = 1
# try: sgd, momentum, rmsprop, adagrad, adadelta, adam, nesterov_momentum
optimizer = "adam"
train_set, valid_set, test_set = data.mnist()
train_xy = data.batched_mnist(train_set, batch_size)
dim_x = train_xy[0][0].shape[1]
dim_y = train_xy[0][1].shape[1]
print "#features = ", dim_x, "#labels = ", dim_y
print "compiling..."
model = AVAE(dim_x, dim_x, hidden_size, latent_size, optimizer)
print "training..."
start = time.time()
for i in xrange(200):
error = 0.0
error_d = 0.0
error_g = 0.0
in_start = time.time()
for batch_id, xy in train_xy.items():
X = xy[0]
if i < 50:
error += model.train_vae(X, lr)
continue
local_bath_size = len(X)
Z = model.noiser(local_bath_size)
loss_d = 0
if batch_id % iter_d == 0:
loss_d += model.train_d(X, Z, lr)
cost, loss_g = model.train_g(X, Z, lr)
error += cost
error_g += loss_g
error_d += loss_d
in_time = time.time() - in_start
error /= len(train_xy);
error_d /= len(train_xy) / iter_d;
error_g /= len(train_xy);
print "Iter = " + str(i) + ", vlbd = " + str(error) \
+ ", error_d = " + str(error_d) + ", error_g = " + str(error_g) \
+ ", Time = " + str(in_time)
print "training finished. Time = " + str(time.time() - start)
print "save model..."
save_model("./model/vae_mnist.model", model)
'''-------------Visualization------------------'''
# code from: https://jmetzen.github.io/2015-11-27/vae.html
load_model("./model/vae_mnist.model", model)
print "validation.."
valid_xy = data.batched_mnist(valid_set, batch_size)
error = 0
for batch_id, xy in valid_xy.items():
X = xy[0]
cost, y = model.validate(X)
error += cost
print "Loss = " + str(error / len(valid_xy))
plt.figure(figsize=(8, 12))
for i in range(5):
plt.subplot(5, 2, 2*i + 1)
plt.imshow(X[i].reshape(28, 28), vmin=0, vmax=1)
plt.title("Test input")
plt.colorbar()
plt.subplot(5, 2, 2*i + 2)
plt.imshow(y[i].reshape(28, 28), vmin=0, vmax=1)
plt.title("Reconstruction")
plt.colorbar()
plt.tight_layout()
plt.savefig("reconstruct.png", bbox_inches="tight")
plt.show()
## manifold
if latent_size == 2:
test_xy = data.batched_mnist(test_set, 5000)
X = test_xy[0][0]
mu = np.array(model.project(X))
plt.figure(figsize=(8, 6))
plt.scatter(mu[:, 0], mu[:, 1], c=np.argmax(np.array(test_xy[0][1]), 1))
plt.colorbar()
plt.savefig("2dstructure.png", bbox_inches="tight")
plt.show()
'''--------------------------'''
nx = ny = 20
x_values = np.linspace(-3, 3, nx)
y_values = np.linspace(-3, 3, ny)
canvas = np.empty((28*ny, 28*nx))
for i, yi in enumerate(x_values):
for j, xi in enumerate(y_values):
z = np.array([[xi, yi]], dtype=theano.config.floatX)
y = model.generate(z)
canvas[(nx-i-1)*28:(nx-i)*28, j*28:(j+1)*28] = y.reshape(28, 28)
fit = plt.figure(figsize=(8, 10))
Xi, Yi = np.meshgrid(x_values, y_values)
plt.imshow(canvas, origin="upper")
plt.tight_layout()
plt.savefig("manifold.png", bbox_inches="tight")
plt.show()