A new primal-dual algorithm for minimizing the sum of three functions with a linear operator

TL;DR

This paper introduces a new primal-dual algorithm for convex optimization involving three functions and a linear operator, extending existing methods with proven convergence and improved efficiency.

Contribution

It presents a novel primal-dual algorithm that generalizes and improves upon existing methods for minimizing the sum of three convex functions with convergence guarantees.

Findings

Proves convergence in terms of fixed point distance.

Establishes an $O(1/k)$ ergodic convergence rate.

Demonstrates efficiency through numerical experiments.

Abstract

In this paper, we propose a new primal-dual algorithm for minimizing $f(x) + g(x) + h(Ax)$, where $f$, $g$, and $h$ are proper lower semi-continuous convex functions, $f$ is differentiable with a Lipschitz continuous gradient, and $A$ is a bounded linear operator. The proposed algorithm has some famous primal-dual algorithms for minimizing the sum of two functions as special cases. E.g., it reduces to the Chambolle-Pock algorithm when $f = 0$ and the proximal alternating predictor-corrector when $g = 0$. For the general convex case, we prove the convergence of this new algorithm in terms of the distance to a fixed point by showing that the iteration is a nonexpansive operator. In addition, we prove the $O(1/k)$ ergodic convergence rate in the primal-dual gap. With additional assumptions, we derive the linear convergence rate in terms of the distance to the fixed point. Comparing to other primal-dual algorithms for solving the same problem, this algorithm extends the range of acceptable parameters to ensure its convergence and has a smaller per-iteration cost. The numerical experiments show the efficiency of this algorithm.