Online Linear Programming: Dual Convergence, New Algorithms, and Regret Bounds

TL;DR

This paper advances online linear programming by proving dual price convergence for general LPs, introducing a new algorithm with improved regret bounds, and validating its effectiveness through experiments.

Contribution

It establishes dual convergence results for general LPs, proposes an action-history-dependent algorithm, and achieves a logarithmic regret bound, extending prior work focused on nonnegative coefficients.

Findings

Dual prices converge to offline LP solutions under regularity conditions.

Proposed algorithm attains an $O( ext{log} n ext{log} ext{log} n)$ regret bound.

Numerical experiments confirm the algorithm's superior performance.

Abstract

We study an online linear programming (OLP) problem under a random input model in which the columns of the constraint matrix along with the corresponding coefficients in the objective function are generated i.i.d. from an unknown distribution and revealed sequentially over time. Virtually all pre-existing online algorithms were based on learning the dual optimal solutions/prices of the linear programs (LP), and their analyses were focused on the aggregate objective value and solving the packing LP where all coefficients in the constraint matrix and objective are nonnegative. However, two major open questions were: (i) Does the set of LP optimal dual prices learned in the pre-existing algorithms converge to those of the "offline" LP, and (ii) Could the results be extended to general LP problems where the coefficients can be either positive or negative. We resolve these two questions by establishing convergence results for the dual prices under moderate regularity conditions for general LP problems. Specifically, we identify an equivalent form of the dual problem which relates the dual LP with a sample average approximation to a stochastic program. Furthermore, we propose a new type of OLP algorithm, Action-History-Dependent Learning Algorithm, which improves the previous algorithm performances by taking into account the past input data as well as decisions/actions already made. We derive an $O(\log n \log \log n)$ regret bound (under a locally strong convexity and smoothness condition) for the proposed algorithm, against the $O(\sqrt{n})$ bound for typical dual-price learning algorithms, where $n$ is the number of decision variables. Numerical experiments demonstrate the effectiveness of the proposed algorithm and the action-history-dependent design.