96
20

Global Riemannian Acceleration in Hyperbolic and Spherical Spaces

Abstract

We further research on the accelerated optimization phenomenon on Riemannian manifolds by introducing accelerated global first-order methods for the optimization of LL-smooth and geodesically convex (g-convex) or μ\mu-strongly g-convex functions defined on the hyperbolic space or a subset of the sphere. For a manifold other than the Euclidean space, these are the first methods to \emph{globally} achieve the same rates as accelerated gradient descent in the Euclidean space with respect to LL and ε\varepsilon (and μ\mu if it applies), up to log factors. Previous results with these accelerated rates only worked, given strong g-convexity, in a generally small neighborhood (initial distance RR to a minimizer being R=O((μ/L)3/4)R = O((\mu/L)^{3/4})). Our rates have a polynomial factor on 1/cos(R)1/\cos(R) (spherical case) or cosh(R)\cosh(R) (hyperbolic case). Thus, we completely match the Euclidean case for a constant initial distance, and for larger RR we incur greater constants due to the geometry. As a proxy for our solution, we solve a constrained non-convex Euclidean problem, under a condition between convexity and \textit{quasar-convexity}, of independent interest. Additionally, for any Riemannian manifold of bounded sectional curvature, we provide reductions from optimization methods for smooth and g-convex functions to methods for smooth and strongly g-convex functions and vice versa.

View on arXiv
Comments on this paper