Adaptive Primal-Dual Hybrid Gradient Methods for Saddle-Point Problems

MaRDI QIDQ6241629zbMATH OpenFDO

Authors Tom Goldstein, Min Li, Xiaoming Yuan, Ernie Esser, Richard G. Baraniuk

Publication date 2 May 2013

Has companion code repository https://github.com/martinResearch/PySparseLP

Mathematics Subject Classification ID

Numerical methods for variational inequalities and related problems (65K15)

Abstract: The Primal-Dual hybrid gradient (PDHG) method is a powerful optimization scheme that breaks complex problems into simple sub-steps. Unfortunately, PDHG methods require the user to choose stepsize parameters, and the speed of convergence is highly sensitive to this choice. We introduce new adaptive PDHG schemes that automatically tune the stepsize parameters for fast convergence without user inputs. We prove rigorous convergence results for our methods, and identify the conditions required for convergence. We also develop practical implementations of adaptive schemes that formally satisfy the convergence requirements. Numerical experiments show that adaptive PDHG methods have advantages over non-adaptive implementations in terms of both efficiency and simplicity for the user.

This page was built for publication: Adaptive Primal-Dual Hybrid Gradient Methods for Saddle-Point Problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6241629)