Continuous Control of Conservatism for Robust Optimization by Adjustable Regret

arxiv(2023)

引用 0|浏览0
暂无评分
摘要
It is commonly recognized that a major issue of robust optimization is the tendency to produce overly conservative solutions. To address this issue, a new regret-based criterion with a single control parameter is proposed and axiomatized to offer smooth control of conservatism in a wide range without tampering with the uncertainty set. This criterion has many appealing analytical properties, such as decreasing conservatism with regard to the control parameter, which makes it a unique choice for fine control of conservatism. Tractability for robust linear programs with this criterion is established by reformulating them into those with the maximin criterion, for which tractable solution schemes and theoretical results are actively developed in the literature. Closed-form solutions are obtained for the robust one-way trading problem with this criterion, leading to a greatly simplified competitive ratio analysis. Numerical experiments are conducted to demonstrate smooth control of conservatism and the effects on revenue and risk.
更多
查看译文
关键词
robust optimization,conservatism,control,adjustable
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要