Neural Networks are famously known as black box models in artificial intelligence, that means that we know little to nothing about what happens inside of them. Lately, there has been a big interest in research and multiple techniques are being developed to try and understand a bit more about what happens inside of these models. In particular, this a method recently developed to to help tackle this problem and it presents an algorithm to formally infer certain properties about the network. The only thing is that it relies on the existence of an approximation of a Neural Network using linear piecewise functions. The aim of this work is to study and research methods to generate this approximation and complement the work done in the previously mentioned thesis.