Hello, sorry for the delay!
By the way, please write on the forum if anything may seem wrong with the tests.
I believe that you may implement the FC Layer using the following snippet:
class FullyConnectedLayer(BaseLayer):
def __init__(self, input_size, output_size, learning_rate):
""" A fully connected layer.
param: input_size (int): dimension n of the input vector
param: output_size (int): dimension m of the output vector
param: learning_rate (float): the learning rate of this layer
"""
# TODO: define the neccesary class variables
pass
def forward(self, x):
""" Compute the foward pass through the layer.
param: x (np.ndarray): input with shape [b, n] where b is the batch size and n is the input size
returns (np.ndarray): result of the forward pass, of shape [b, m] where b is the batch size and
m is the output size
"""
# TODO: Implement forward pass of the fully connected layer
# Hint: Think about what you need to store during the forward pass to be able to compute
# the gradients in the backward pass
pass
def get_gradient_weights(self):
"""
returns (np.ndarray): the gradient with respect to the weights and biases from the last call of backward(...)
"""
# TODO: Implement
pass
def backward(self, error):
""" Update the weights of this layer and return the gradient with respect to the previous layer.
param: error (np.ndarray): of shape [b, m] where b is the batch size and m is the output size
returns (np.ndarray): the gradient w.r.t. the previous layer, of shape [b, n] where b is the
batch size and n is the input size
"""
# TODO: Implement backward pass of the fully connected layer
# Hint: Be careful about the order of applying the update to the weights and the calculation of
# the error with respect to the previous layer.
pass
def initialize(self, weights_initializer, bias_initializer):
""" Initializes the weights/bias of this layer with the given initializers.
param: weights_initializer: object providing a method weights_initializer.initialize(weights_shape)
which will return initialized weights with the given shape
param: bias_initializer: object providing a method bias_initializer.initialize(bias_shape)
which will return an initialized bias with the given shape
"""
# TODO: Implement
pass
