Project for the course "Machine Learning" of the University of Bologna, A.Y. 2021/2022.
The purpose of the project is to learn the mapping from polar coordinates to a a discrete 10x10 grid of cells in the plane, using a neural network.
The supervised dataset is given to you in the form of a generator (to be considered as a black box).
The model must achieve an accuracy of 95%, and it will be evaluated in a way inversely proportional to the number of its parameters: the smaller, the better.
WARNING: Any solution taking advantage of meta-knowledge about the generator will be automatically rejected.