This code was made to practice the implementation of normalizing flows using the Real NVP model. This model is one of the tools used when dealing with Quantum Mechanical systems, and a major tool needed for my internship at DESY.
These are also predictions made for two different datasets. The first one is the make_circles dataset, while the second is make_moons. While these distributions are not trivial to describe using two variables, we can see that the model makes a solid approximation. Noise had to be incorporated in order to make it work properly, and different epoch number will lead to different degrees of accuracy that do not scale linearly (or proportionally even).
This third picture is a 3-dimensional prediction made by the same model, which was mainly made to verify the validity for different dimensional data. Noise was also incorporated, and a bigger data sample was used.