Data driven research has becoming an important research line in operations research and operations management. In traditional supply chain management, decisions are often based on strong assumptions of customer demand distribution, for example in classical inventory models the demand are often assumed to follow i.i.d. normal or log-normal distribution, and in the modern online-learning models the underlying assumption is one can learn to as accurate as possible as time passes. However, data in supply chain management and other practical areas are often noisy and biased, and influenced by non-observable or non-predictable factors and events, e.g., weather or regulation changes.
Robust optimization provides a mathematical toolbox for decision under partial distribution information and against uncertainty factors. In this talk, we introduce the basic models and principles for distributional robust optimization, which focuses on uncertainty of the distribution shape and parameters. Recent theoretical progresses and applications are also discussed, in supply chain management and machine learning areas.
何斯迈毕业于中国科学技术大学数学系,在香港中文大学获得运筹学博士学位。现任上海财经大学教授。在Operations Research, Mathematics of Operations Research, Mathematical Programming和SIAM Journal on Optimization上发表论文10余篇。曾获2014年度中国运筹协会青年科技奖及上海市特聘教授(东方学者)称号。2018年获得国家自然科学基金杰出青年资助。