Journal of Machine Learning Research, Volume 12, July 2011, pages 2121--2159. Originally UC Berkeley EECS Technical Report 2010-24.
Conference version. In the International Conference on Learning Theory (COLT 2010)
Slides. Presented at International Symposium on Mathematical Programming (ISMP) 2012
We present a new family of subgradient methods that dynamically incorporate knowledge of the geometry of the data observed in earlier iterations to perform more informative gradient-based learning. Metaphorically, the adaptation allows us to find needles in haystacks in the form of very predictive but rarely seen features. Our paradigm stems from recent advances in stochastic optimization and online learning which employ proximal functions to control the gradient steps of the algorithm. We describe and analyze an apparatus for adaptively modifying the proximal function, which significantly simplifies setting a learning rate and results in regret guarantees that are provably as good as the best proximal function that can be chosen in hindsight. We give several efficient algorithms for empirical risk minimization problems with common and important regularization functions and domain constraints. We experimentally study our theoretical analysis and show that adaptive subgradient methods significantly outperform state-of-the-art, yet non-adaptive, subgradient algorithms.