2026年第60期(总第1204期)
演讲主题:Facet-Defining Generalized Analytical Benders Cuts for Stochastic Mixed-Integer Programming
主讲人:张玉利 北京理工大学管理学院教授
主持人:秦虎 信息管理与数据科学系教授
活动时间:2026年07月11日(周六)10:30-12:00
活动地址:管院大楼105教室
主讲人简介:
张玉利,北京理工大学管理学院教授。研究方向为运筹优化、供应链管理等。国家级人才计划入选者。主持多项国家自然科学基金项目。在MS、POM、IJOC等期刊发表论文。担任北京运筹学会常务理事、中国运筹学会不确定系统分会常务理事、CII、AJOR、运筹与管理等期刊副主编、编委等。曾获INFORMS TSL和INFORMS ENRE Energy、INFORMS SS、华人学者管理科学与工程国际年会等最佳论文奖、华为供应链运筹管理专家委员会合作创新奖等荣誉。
活动简介:
We develop generalized analytical Benders (GAB) cuts for two-stage stochastic programming with mixed-integer recourse. The non-convex recourse function poses challenges for standard decomposition methods. Our GAB cuts are derived from inference duality and branch-and-bound leaf-node information, requiring only one MIP solution and polynomial-time construction of facet-defining linear inequalities. We design an improved logic-based Benders decomposition algorithm that integrates these cuts within a branch-and-cut framework. Notably, the cuts remain valid even when subproblems are solved approximately, enabling significant computational savings. The approach extends to problems with integer first-stage decisions and two-stage robust optimization.