Efficient Online and Batch Learning Using Forward Backward Splitting

Abstract

We describe, analyze, and experiment with a framework for empirical loss minimization with regularization. Our algorithmic framework alternates between two phases. On each iteration we first perfor...