Geom-SPIDER-EM: Faster Variance Reduced Stochastic Expectation Maximization for Nonconvex Finite-Sum Optimization
This work aims to reduce the computational burden of EM for practitioners working with large-scale latent variable models, representing an incremental improvement over existing stochastic EM methods.
This paper addresses the high computational cost of the Expectation Maximization (EM) algorithm in large-scale learning. It proposes Geom-SPIDER-EM, an extension of SPIDER-EM, which achieves state-of-the-art complexity bounds for smooth nonconvex finite-sum optimization problems.
The Expectation Maximization (EM) algorithm is a key reference for inference in latent variable models; unfortunately, its computational cost is prohibitive in the large scale learning setting. In this paper, we propose an extension of the Stochastic Path-Integrated Differential EstimatoR EM (SPIDER-EM) and derive complexity bounds for this novel algorithm, designed to solve smooth nonconvex finite-sum optimization problems. We show that it reaches the same state of the art complexity bounds as SPIDER-EM; and provide conditions for a linear rate of convergence. Numerical results support our findings.