def test_backward(self):
layer = Sigmoid()
x = np.random.rand(2)
y = layer.forward(x)
deriv_grad = layer.backward(np.ones(1))
numerical_grad_matrix = numerical_gradient.calc(layer.forward, x)
# the numerical grad in this case is a matrix made of zeros with
# dJ/dx_i only in the diagonal
num_grad = np.diagonal(numerical_grad_matrix)
numerical_gradient.assert_are_similar(deriv_grad, num_grad)
layer1 = Dense(trainingData.shape[2], 16)
activation1 = Sigmoid()
layer2 = Dense(16, 10)
activation2 = SoftMax()
cost = CostMeanSquared()
for epoch in range(EPOCHS):
print('Epoch: ' + str(epoch + 1) + '/' + str(EPOCHS))
print('')
correct = 0
for batch in range(total // BATCH_SIZE):
### SOCHASIC GRADIENT DESCENT ###
layer1.forward(trainingData[batch])
activation1.forward(layer1.outputs)
layer2.forward(activation1.outputs)
activation2.forward(layer2.outputs)
cost.forward(activation2.outputs, labels[batch], 10)
for sample in range(activation2.outputs.shape[1]):
if np.argmax(activation2.outputs[:, sample]) == np.argmax(
labels[batch, sample]):
correct += 1
cost.backward(activation2.outputs, labels[batch], 10)
activation2.backward(layer2.outputs, layer2.weights.shape[0],
BATCH_SIZE)
layer2.backward(activation1.outputs)
activation1.backward(layer1.outputs)
layer1.backward(trainingData[batch])
def test_forward(self):
layer = Sigmoid()
x = np.array([2., 3., 4.])
y = layer.forward(x)
import numpy as np
try:
sys.path.append(os.path.join(Path(os.getcwd()).parent, 'lib'))
from layers import Sigmoid
except ImportError:
print('Library Module Can Not Found')
# Test1(Vector)
layer = Sigmoid()
x = np.array([0.1, -0.2, 0.3, -0.4, 0.5])
print(x)
y = layer.forward(x)
print(y)
print(layer.out)
dout = np.array([-0.1, -0.2, -0.3, 0.4, -0.5])
dout = layer.backward(dout)
print(dout)
print('=========================================')
# Test2(Matrix)
x = np.array([
[0.1, -0.5, 1.0],
[0.2, -0.6, 2.0],
[0.3, -0.7, 3.0],
[0.4, -0.8, 4.0]