Do landscape analyses suffice for understanding optimization in deep learning?
Off the Convex Path,
Neural network optimization is fundamentally non-convex, and yet simple gradient-based algorithms seem to consistently solve…
Neural network optimization is fundamentally non-convex, and yet simple gradient-based algorithms seem to consistently solve…
A core, emerging problem in nonconvex optimization involves the escape of saddle points. While recent research has shown that…
Vlad Voroninski recently posted an arXiv preprint with Paul Hand that provides compressed sensing guarantees using a neural net-b…