def forward(self, inputs, params):
outputs, running_mean, running_variance = layers.batchnorm(
inputs, params[self.gamma], params[self.beta],
params['__training_in_progress__'], self.epsilon, self.momentum,
self.running_mean, self.running_variance)
self.running_mean, self.running_variance = running_mean, running_variance
return outputs
def forward(self, inputs, params):
outputs, running_mean, running_variance = layers.batchnorm(
inputs, params[self.gamma], params[self.beta],
params['__training_in_progress__'], self.epsilon, self.momentum,
self.running_mean, self.running_variance)
self.running_mean, self.running_variance = running_mean, running_variance
return outputs
def forward(self, inputs, params):
N, C, W, H = inputs.shape
inputs = transpose(inputs, (0, 2, 3, 1))
inputs = np.reshape(inputs, (N * W * H, C))
outputs, running_mean, running_variance = layers.batchnorm(
inputs, params[self.gamma], params[self.beta],
params['__training_in_progress__'], self.epsilon, self.momentum,
self.running_mean, self.running_variance)
self.running_mean, self.running_variance = running_mean, running_variance
outputs = np.reshape(outputs, (N, W, H, C))
outputs = transpose(outputs, (0, 3, 1, 2))
return outputs
def forward(self, inputs, params):
N, C, W, H = inputs.shape
inputs = transpose(inputs, (0, 2, 3, 1))
inputs = np.reshape(inputs, (N * W * H, C))
outputs, running_mean, running_variance = layers.batchnorm(
inputs, params[self.gamma], params[self.beta],
params['__training_in_progress__'], self.epsilon, self.momentum,
self.running_mean, self.running_variance)
self.running_mean, self.running_variance = running_mean, running_variance
outputs = np.reshape(outputs, (N, W, H, C))
outputs = transpose(outputs, (0, 3, 1, 2))
return outputs
def forward(self, X, mode):
# Flatten the input data to matrix.
X = np.reshape(X, (batch_size, 3 * 32 * 32))
# First affine layer (fully-connected layer).
y1 = layers.affine(X, self.params['w1'], self.params['b1'])
# ReLU activation.
y2 = layers.relu(y1)
# Batch normalization
y3, self.aux_params['running_mean'], self.aux_params['running_var'] = layers.batchnorm(
y2, self.params['gamma'], self.params['beta'], running_mean=self.aux_params['running_mean'], \
running_var=self.aux_params['running_var'])
# Second affine layer.
y4 = layers.affine(y3, self.params['w2'], self.params['b2'])
# Dropout
y5 = layers.dropout(y4, 0.5, mode=mode)
return y5
def forward(self, X, mode):
# Flatten the input data to matrix.
X = np.reshape(X, (batch_size, 3 * 32 * 32))
# First affine layer (fully-connected layer).
y1 = layers.affine(X, self.params['w1'], self.params['b1'])
# ReLU activation.
y2 = layers.relu(y1)
# Batch normalization
y3, self.aux_params['running_mean'], self.aux_params['running_var'] = layers.batchnorm(
y2, self.params['gamma'], self.params['beta'], running_mean=self.aux_params['running_mean'], \
running_var=self.aux_params['running_var'])
# Second affine layer.
y4 = layers.affine(y3, self.params['w2'], self.params['b2'])
# Dropout
y5 = layers.dropout(y4, 0.5, mode=mode)
return y5
def check_beta(b):
y, _, _ = layers.batchnorm(x, gamma, b)
return layers.l2_loss(y, fake_y)
def check_gamma(g):
y, _, _ = layers.batchnorm(x, g, beta)
return layers.l2_loss(y, fake_y)
def check_beta(b):
y, _, _ = layers.batchnorm(x, gamma, b)
return layers.softmax_loss(y, fake_y)
def check_gamma(g):
y, _, _ = layers.batchnorm(x, g, beta)
return layers.softmax_loss(y, fake_y)