Timezone: »

Sub-sampled Cubic Regularization for Non-convex Optimization
Jonas Kohler · Aurelien Lucchi

Mon Aug 07 10:30 PM -- 10:48 PM (PDT) @ Parkside 2

We consider the minimization of non-convex functions that typically arise in machine learning. Specifically, we focus our attention on a variant of trust region methods known as cubic regularization. This approach is particularly attractive because it escapes strict saddle points and it provides stronger convergence guarantees than first- and second-order as well as classical trust region methods. However, it suffers from a high computational complexity that makes it impractical for large-scale learning. Here, we propose a novel method that uses sub-sampling to lower this computational cost. By the use of concentration inequalities we provide a sampling scheme that gives sufficiently accurate gradient and Hessian approximations to retain the strong global and local convergence guarantees of cubically regularized methods. To the best of our knowledge this is the first work that gives global convergence guarantees for a sub-sampled variant of cubic regularization on non-convex functions. Furthermore, we provide experimental results supporting our theory.

Author Information

Jonas Kohler (ETH Zurich)
Aurelien Lucchi (ETH)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors