RL Simplex: Bringing Computational Efficiency in Linear Programming via Reinforcement Learning
Primary Area: optimization
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Keywords: Reinforcement learning, Pivot rules, Simplex method, Linear programming, TSP
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
Abstract: In the simplex method, the selection of variables during the pivot operation in each iteration significantly impacts the overall computational process. The primary objective of this study is to provide explicit guidance for the selection of pivot variables, particularly when multiple candidate variables for pivoting are available, through the application of reinforcement learning techniques. We illustrate our approach, termed RL Simplex, to the Euclidean Traveling Salesman Problem (TSP) with varying city counts, substantially reducing the number of iterations. Our experimental findings demonstrate the practical feasibility and successful integration of reinforcement learning with the simplex method, surpassing the performance of established solver software packages such as Gurobi and SciPy.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 5746
Loading