A Decentralized Primal Dual Algorithm with Quasi-Newton Tracking

arXiv2304.01614MaRDI QIDQ6509495FDOQ6509495

Authors: Hao Wu, Li-ping Wang, Hongchao Zhang 

Abstract: In this paper, we consider the decentralized optimization problem of minimizing a finite sum of strongly convex and twice continuously differentiable functions over a fixed connected undirected network. We propose a fully decentralized second-order multiplier algorithm(DSOMM) where second-order information are approximated by simple algebraic calculations and at most matrix-vector products. It is the first to study the second-order multiplier method in decentralized optimization where we find this complex method is equivalent to the primal-dual method with twice primal quasi-Newton steps and once dual BB step per iteration. Additionally, in DSOMM the local direction on each node asymptotically approximates the global quasi-Newton direction. Under some mild assumptions, the proposed algorithm is showed to have global linear convergence rate. Numerical results are also reported for verifying the effectiveness.

This page was built for publication: A Decentralized Primal Dual Algorithm with Quasi-Newton Tracking

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6509495)