.. _`sec:back-prop`:

Back-propagation
----------------

To find the optimum weights and biaises of a FNN, the back-propagation is used. 
The gradient of the loss function with respect to these parameters (i.e. weights and biases) has to be calculated. 
This stochastic gradient descent is then used in an optimization algorithm. 
The goal of this section is to show how this gradient is obtained. 


.. toctree::
   
   backprop_simple_case
   backprop_real_case