no code implementations • ICML 2018 • Thomas Laurent, James Brecht
We consider deep linear networks with arbitrary convex differentiable loss.