Nonconvex learning via Stochastic Gradient Langevin Dynamics: a nonasymptotic analysis
[edit]
Proceedings of the 2017 Conference on Learning Theory, PMLR 65:16741703, 2017.
Abstract
Stochastic Gradient Langevin Dynamics (SGLD) is a popular variant of Stochastic Gradient Descent, where properly scaled isotropic Gaussian noise is added to an unbiased estimate of the gradient at each iteration. This modest change allows SGLD to escape local minima and suffices to guarantee asymptotic convergence to global minimizers for sufficiently regular nonconvex objectives. The present work provides a nonasymptotic analysis in the context of nonconvex learning problems, giving finitetime guarantees for SGLD to find approximate minimizers of both empirical and population risks. As in the asymptotic setting, our analysis relates the discretetime SGLD Markov chain to a continuoustime diffusion process. A new tool that drives the results is the use of weighted transportation cost inequalities to quantify the rate of convergence of SGLD to a stationary distribution in the Euclidean $2$Wasserstein distance.
Related Material


