phitodeep.layers.base
=====================

.. py:module:: phitodeep.layers.base


Classes
-------

.. autoapisummary::

   phitodeep.layers.base.Layer
   phitodeep.layers.base.Flatten
   phitodeep.layers.base.Dense


Module Contents
---------------

.. py:class:: Layer(name: str, initializer: phitodeep.optimization.initialization.Initializer = Initializer())

   Base class for all layers in the network.


   .. py:attribute:: name


   .. py:attribute:: cache


   .. py:attribute:: grads


   .. py:attribute:: initializer


   .. py:method:: forward(X: numpy.ndarray)
      :abstractmethod:



   .. py:method:: backward(dL_dZ)
      :abstractmethod:


      Backward pass through the block.

      :param dL_dZ: gradient of loss w.r.t. output of this block

      :returns: gradient of loss w.r.t. input (to pass to previous layer)
      :rtype: dL_dX



   .. py:method:: copy()
      :abstractmethod:



.. py:class:: Flatten

   Bases: :py:obj:`Layer`


   Flattens the input tensor into a 2D tensor.


   .. py:method:: forward(X: numpy.ndarray)

      X: (batch_size, ...) -> (batch_size, ...)



   .. py:method:: backward(dL_dZ)

      dL_dZ: (batch_size, ...) -> (batch_size, ...)



   .. py:method:: copy()


.. py:class:: Dense(input_size: int, output_size: int, initializer: phitodeep.optimization.initialization.Initializer = He())

   Bases: :py:obj:`Layer`


   Fully connected layer.


   .. py:attribute:: grads


   .. py:attribute:: input_size


   .. py:attribute:: output_size


   .. py:attribute:: W


   .. py:attribute:: b


   .. py:method:: forward(X: numpy.ndarray)

      X: (batch_size, input_size) -> (batch_size, output_size)



   .. py:method:: backward(dL_dZ)

      Backpropagate through Dense layer.

      :param dL_dZ: (batch_size, output_size) - gradient of loss w.r.t. output

      :returns: (batch_size, input_size) - gradient to pass to previous layer
      :rtype: dL_dX



   .. py:method:: copy()