Academic Journal
Beyond O(T) Regret: Decoupling Learning and Decision Making in Online Linear Programming.
| Τίτλος: | Beyond O(T) Regret: Decoupling Learning and Decision Making in Online Linear Programming. |
|---|---|
| Συγγραφείς: | Gao, Wenzhi1 (AUTHOR) gwz@stanford.edu, Ge, Dongdong2 (AUTHOR) ddge@sjtu.edu.cn, Sun, Chunlin1 (AUTHOR) chunlin@stanford.edu, Xue, Chenyu3 (AUTHOR) xcy2721d@gmail.com, Ye, Yinyu2 (AUTHOR) yinyuye@sjtu.edu.cn |
| Πηγή: | Operations Research. Jul/Aug2026, Vol. 74 Issue 4, p1932-1944. 13p. |
| Θεματικοί όροι: | *Linear programming, *Decision making, *Scalability, Subgradient methods, Duality theory (Mathematics) |
| Περίληψη: | How can fast first-order methods for online linear programming (OLP) match the performance of more computationally intensive LP-based approaches? In the paper "Beyond O(T) Regret: Decoupling Learning and Decision Making in Online Linear Programming," the authors develop a new framework that improves the regret guarantees of scalable first-order algorithms for online linear programming. They show that when the dual problem satisfies a mild error bound condition, first-order methods can achieve o(T) regret, including O(T1/3) in continuous support settings and O(log T) in finite support settings. The key insight is to decouple learning and decision making. The algorithm first learns an approximate dual solution and then localizes decisions to a neighborhood around that solution using an adaptive procedure. This decoupling narrows the gap between efficient gradient-based methods and LP-based benchmarks, providing strong theoretical guarantees while maintaining scalability for applications such as revenue management, online advertising, and cloud computing. Online linear programming plays an important role in both revenue management and resource allocation, and recent research has focused on developing efficient first-order online learning algorithms. Despite the empirical success of first-order methods, they typically achieve a regret no better than O(T) , which is suboptimal compared with the O(log T) bound guaranteed by the state-of-the-art linear programming (LP)-based online algorithms. This paper establishes a general framework that improves on the O(T) result when the LP dual problem exhibits certain error bound conditions. For the first time, we show that first-order learning algorithms achieve o(T) regret in the continuous support setting and O(log T) regret in the finite support setting beyond the nondegeneracy assumption. Our results significantly improve the state-of-the-art regret results and provide new insights for sequential decision making. Funding: This research was supported by the National Natural Science Foundation of China [Grants 72225009, 72394360, and 72394365]. Supplemental Material: All supplemental materials, including the code, data, and files required to reproduce the results, are available at https://doi.org/10.1287/opre.2024.1575. [ABSTRACT FROM AUTHOR] |
| Copyright of Operations Research is the property of INFORMS: Institute for Operations Research & the Management Sciences and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Βάση Δεδομένων: | Business Source Index |
καταχωρήστε σχόλιο πρώτοι!