Faster First-Order Methods for Stochastic Non-Convex Optimization on Riemannian Manifolds

First-order non-convex Riemannian optimization algorithms have gained recent popularity in structured machine learning problems including principal component analysis and low-rank matrix completion. The current paper presents an efficient Riemannian Stochastic Path Integrated Differential EstimatoR (R-SPIDER) algorithm to solve the finite-sum and online Riemannian non-convex minimization problems. At the core of R-SPIDER is a recursive semi-stochastic gradient estimator that can accurately estimate Riemannian gradient under not only exponential mapping and parallel transport, but also general retraction and vector transport operations. Compared with prior Riemannian algorithms, such a recursive gradient estimation mechanism endows R-SPIDER with lower computational cost in first-order oracle complexity. Specifically, for finite-sum problems with n components, R-SPIDER is proved to converge to an epsilon-approximate stationary point within O(min(n + root n/epsilon(2),1/epsilon(3))) stochastic gradient evaluations, beating the best-known complexity O(n+1/epsilon(4)); for online optimization, R-SPIDER is shown to converge with O(1/epsilon(3)) complexity which is, to the best of our knowledge, the first non-asymptotic result for online Riemannian optimization. For the special case of gradient dominated functions, we further develop a variant of R-SPIDER with improved linear rate of convergence. Extensive experimental results demonstrate the advantage of the proposed algorithms over the state-of-the-art Riemannian non-convex optimization methods.

Automated summary

The paper introduces an efficient Riemannian Stochastic Path Integrated Differential Estimator algorithm for solving finite and online Riemannian non-convex minimization problems with lower computational cost compared to prior algorithms. The algorithm's recursive gradient estimation mechanism allows it to achieve epsilon-approximate stationary points within a certain number of stochastic gradient evaluations, outperforming other Riemannian optimization methods.