Training an algorithm - generic

[martinapugliese]

There should be a generic intro to ML and how to train an algorithm, with a reference to gradient descent which is in the maths folder and not super visible there.

[martinapugliese]

Add a section on cost functions and refer to the gradient descent indeed. For instance the chapter on backpropagation in neural networks should also point there.

[martinapugliese]

There is now a subsection for general ML, add there

[martinapugliese]

Add training/testing on the same data and in-sample/out-of-sample errors