A New Crossover Algorithm for LP Inspired by the Spiral Dynamic of PDHG

Tianhao Liu,Haihao Lu

2024-09-23

Abstract:Motivated by large-scale applications, there is a recent trend of research on using first-order methods for solving LP. Among them, PDLP, which is based on a primal-dual hybrid gradient (PDHG) algorithm, may be the most promising one. In this paper, we present a geometric viewpoint on the behavior of PDHG for LP. We demonstrate that PDHG iterates exhibit a spiral pattern with a closed-form solution when the variable basis remains unchanged. This spiral pattern consists of two orthogonal components: rotation and forward movement, where rotation improves primal and dual feasibility, while forward movement advances the duality gap. We also characterize the different situations in which basis change events occur. Inspired by the spiral behavior of PDHG, we design a new crossover algorithm to obtain a vertex solution from any optimal LP solution. This approach differs from traditional simplex-based crossover methods. Our numerical experiments demonstrate the effectiveness of the proposed algorithm, showcasing its potential as an alternative option for crossover.

Optimization and Control

What problem does this paper attempt to address?

The main issues that this paper attempts to address in linear programming (LP) are two key aspects:

1. **Analyzing the behavior of the PDHG algorithm in solving linear programming**:
- The paper first provides a geometric perspective to understand the convergence behavior of the primal-dual hybrid gradient (PDHG) method when solving linear programming. The study finds that the PDHG iterative process exhibits a spiral pattern, consisting of two orthogonal components: rotation and translation. The rotation improves the feasibility of the primal and dual, while the translation reduces the dual gap.
- The study also analyzes the conditions under which basis change events occur in different scenarios.

2. **Designing a new crossover algorithm to obtain vertex solutions from any optimal solution**:
- Traditionally, converting the optimal solution obtained from interior-point methods to a vertex solution requires the use of the simplex method for crossover processing. This paper proposes a new crossover algorithm inspired by the spiral behavior of PDHG, which generates vertex solutions from PDLP solutions without relying on simplex steps.
- The new algorithm leverages the PDLP itself and its spiral axis (referred to as the spiral ray), avoiding the use of the simplex method, and its effectiveness is validated through numerical experiments.

In summary, the paper aims to deeply understand the behavioral characteristics of the PDHG algorithm in solving linear programming problems and, based on this, proposes a novel and efficient crossover algorithm to transform any optimal solution into a vertex solution. This provides new ideas and technical means for solving large-scale linear programming problems.