腾讯 AI Lab 官网

腾讯 AI Lab 官网
Adaptive Proximal Average Approximation for Composite Convex Minimization
Abstract View the Paper
We propose a fast first-order method to solve multi-term nonsmooth composite convex minimization problems by employing a recent proximal average approximation technique and a novel adaptive parameter tuning technique. Thanks to this powerful parameter tuning technique, the proximal gradient step can be performed with a much larger stepsize in the algorithm implementation compared with the prior PA-APG method, which is the core to enable significant improvements in practical performance. Moreover, by choosing the approximation parameter adaptively, the proposed method is shown to enjoy the O(1/k) iteration complexity theoretically without needing any extra computational cost, while the PA-APG method incurs much more iterations for convergence. The preliminary experimental results on overlapping group Lasso and graph-guided fused Lasso problems confirm our theoretic claim well, and indicate that the proposed method is almost five times faster than the state-of-the-art PA-APG method and therefore suitable for higher-precision required optimization.
Venue
In Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence (AAAI 2017), San Francisco, California, USA
Publication Time
Feb 2017
Authors
Li Shen, Wei Liu, Junzhou Huang, Yu-Gang Jiang, and Shiqian Ma