Project realized as part of a course of Stochastic Optimization and Automatic Differentiation for Machine Learning at ENSAE. Report in French.

I used automatic differenciation to train a neural network to predict in which part of the world a given piece of music has been created based on its caracteristics. I trained 2 models, one giving a prediction of the geographical coordinates (latitude & longitude) and another one predicting a heatmap giving density probabilities. I made some nice maps to visualize the predictions, go check them !

I used data from the Geographical Origin of Music dataset provided on the UCI Machine Learning Repository : https://archive.ics.uci.edu/ml/datasets/Geographical+Original+of+Music