Generalization emerges from local optimization in a self-organized learning network

S. Barland, L. Gil

We design and analyze a new paradigm for building supervised learning networks, driven only by local optimization rules without relying on a global error function. Traditional neur…