class TestRelu(unittest.TestCase):
def setUp(self):
self.relu = Relu()
def test_forward(self):
x = np.array([[1.0, -0.5], [-2.0, 3.0]])
assert_array_equal(np.array([[1., 0.], [0., 3.]]),
self.relu.forward(x))
def test_backward(self):
x = np.array([[1.0, -0.5], [-2.0, 3.0]])
assert_array_equal(np.array([[1., 0.], [0., 3.]]),
self.relu.backward(self.relu.forward(x)))
# training
for i in range(int(images.shape[0] / batch_size)):
# forward
img = images[i * batch_size:(i + 1) * batch_size].reshape(
[batch_size, 28, 28, 1])
label = labels[i * batch_size:(i + 1) * batch_size]
conv1_out = conv1.forward(img)
relu1_out = relu1.forward(conv1_out)
pool1_out = pool1.forward(relu1_out)
conv2_out = conv2.forward(pool1_out)
relu2_out = relu2.forward(conv2_out)
pool2_out = pool2.forward(relu2_out)
fc_out = fc.forward(pool2_out)
print("loss: %10.3f" % sf.cal_loss(fc_out, np.array(label)))
# backward
sf.backward()
fc_back = fc.backward(sf.eta)
pool2_back = pool2.backward(fc_back)
relu2_back = relu2.backward(pool2_back)
conv2_back = conv2.backward(relu2_back)
pool1_back = pool1.backward(conv2_back)
relu1_back = relu1.backward(pool1_back)
conv1_back = conv1.backward(relu1_back)
if i % 1 == 0:
fc.gradient(alpha=learning_rate, weight_decay=0.0004)
conv2.gradient(alpha=learning_rate, weight_decay=0.0004)
conv1.gradient(alpha=learning_rate, weight_decay=0.0004)