forked from oreilly-japan/deep-learning-from-scratch-3
-
Notifications
You must be signed in to change notification settings - Fork 93
/
Copy pathtest_relu.py
59 lines (47 loc) · 1.87 KB
/
test_relu.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
import unittest
import numpy as np
from dezero import Variable
import dezero.functions as F
from dezero.utils import gradient_check, array_allclose, array_equal
import chainer.functions as CF
class TestRelu(unittest.TestCase):
def test_forward1(self):
x = np.array([[-1, 0], [2, -3], [-2, 1]], np.float32)
res = F.relu(x)
ans = np.array([[0, 0], [2, 0], [0, 1]], np.float32)
self.assertTrue(array_allclose(res, ans))
def test_backward1(self):
x_data = np.array([[-1, 1, 2], [-1, 2, 4]])
self.assertTrue(gradient_check(F.relu, x_data))
def test_backward2(self):
np.random.seed(0)
x_data = np.random.rand(10, 10) * 100
self.assertTrue(gradient_check(F.relu, x_data))
def test_backward3(self):
np.random.seed(0)
x_data = np.random.rand(10, 10, 10) * 100
self.assertTrue(gradient_check(F.relu, x_data))
class TestLeakyRelu(unittest.TestCase):
def test_forward1(self):
x = np.array([[-1, 0], [2, -3], [-2, 1]], np.float32)
res = F.leaky_relu(x)
ans = np.array([[-0.2, 0.], [2., -0.6], [-0.4, 1.]], np.float32)
self.assertTrue(array_allclose(res, ans))
def test_forward2(self):
slope = 0.002
x = np.random.randn(100)
y2 = CF.leaky_relu(x, slope)
y = F.leaky_relu(x, slope)
res = array_allclose(y.data, y2.data)
self.assertTrue(res)
def test_backward1(self):
x_data = np.array([[-1, 1, 2], [-1, 2, 4]])
self.assertTrue(gradient_check(F.leaky_relu, x_data))
def test_backward2(self):
np.random.seed(0)
x_data = np.random.rand(10, 10) * 100
self.assertTrue(gradient_check(F.leaky_relu, x_data))
def test_backward3(self):
np.random.seed(0)
x_data = np.random.rand(10, 10, 10) * 100
self.assertTrue(gradient_check(F.leaky_relu, x_data))