Abstract: We propose a parallel stochastic Newton method (PSN) for minimizing unconstrained smooth convex functions. We analyze the method in the strongly convex case, and give conditions under which acceleration can be expected when compared to its serial counterpart. We show how PSN can be applied to the empirical risk minimization problem, and demonstrate the practical efficiency of the method through numerical experiments and models of simple matrix classes.
Recommendations
- A parallel inexact Newton method for stochastic programs with recourse
- An asynchronous parallel Newton method
- Sub-sampled Newton methods
- On the asymptotic rate of convergence of stochastic Newton algorithms and their weighted averaged versions
- An Asynchronous Parallel Stochastic Coordinate Descent Algorithm
Cited in
(7)- An asynchronous parallel Newton method
- On the local convergence of a stochastic semismooth Newton method for nonsmooth nonconvex optimization
- Use of projective coordinate descent in the Fekete problem
- Discriminative Bayesian filtering lends momentum to the stochastic Newton method for minimizing log-convex functions
- On the asymptotic rate of convergence of stochastic Newton algorithms and their weighted averaged versions
- Parallel algorithms and probability of large deviation for stochastic convex optimization problems
- Stochastic reformulations of linear systems: algorithms and convergence theory
This page was built for publication: Parallel stochastic Newton method
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4688182)
