This repository contains the codes to reproduce the experiments in the paper "Learning Dynamics of Direct Feedback Alignment", by Maria Refinetti, Stéphane d'Ascoli, Ruben Ohana and Sebastian Goldt.

It is separated into two folders : the one named "shallow" contains experiments on shallow networks in the online setup, whereas the one named "deep" contains experiments on deep networks in the general setup. Both contain their own README files.