Skip to yearly menu bar Skip to main content


Poster

Inexact trust-region algorithms on Riemannian manifolds

Hiroyuki Kasai · Bamdev Mishra

Room 210 #15

Keywords: [ Nonlinear Dimensionality Reduction and Manifold Learning ] [ Non-Convex Optimization ]


Abstract:

We consider an inexact variant of the popular Riemannian trust-region algorithm for structured big-data minimization problems. The proposed algorithm approximates the gradient and the Hessian in addition to the solution of a trust-region sub-problem. Addressing large-scale finite-sum problems, we specifically propose sub-sampled algorithms with a fixed bound on sub-sampled Hessian and gradient sizes, where the gradient and Hessian are computed by a random sampling technique. Numerical evaluations demonstrate that the proposed algorithms outperform state-of-the-art Riemannian deterministic and stochastic gradient algorithms across different applications.

Live content is unavailable. Log in and register to view live content