Overshoot: Taking advantage of future gradients in momentum-based stochastic optimization

Open in new window