Adaptive Proximal Gradient Methods for Structured Neural Networks

Lastly, we demonstrate the superiority of stochastic proximal methods compared to subgradient-based approaches via extensive experiments.