Langevin Dynamics with Continuous Tempering for High-dimensional Non-convex Optimization.
Accepted version
Repository URI
Repository DOI
Change log
Authors
Ye, Nanyang
Zhu, Zhanxing
Mantiuk, Rafal K
Abstract
Minimizing non-convex and high-dimensional objective functions is challenging, especially when training modern deep neural networks. In this paper, a novel approach is proposed which divides the training process into two consecutive phases to obtain better generalization performance: Bayesian sampling and stochastic optimization. The first phase is to explore the energy landscape and to capture the"temperature dynamics''. These strategies can overcome the challenge of early trapping into bad local minima and have achieved remarkable improvements in various types of neural networks as shown in our theoretical analysis and empirical experiments.
Description
Keywords
Journal Title
CoRR
Conference Name
Journal ISSN
Volume Title
abs/1703.04379