authornorthwesterndrostochastic-optimizationsemi-infinite-programming

Sanjay Mehrotra

Department of Industrial Engineering and Management Sciences, Northwestern University, Evanston, IL 60208. Email: mehrotra@northwestern.edu.

Sanjay Mehrotra is a professor of Industrial Engineering and Management Sciences at Northwestern University, with a research programme spanning distributionally robust optimisation, stochastic programming, semi-infinite programming, and statistical-learning regularisation. He is the senior author of the comprehensive 2022 survey Distributionally Robust Optimization: A Review (with Hamed Rahimian, his then-PhD student) and a long-standing contributor to the algorithmic side of DRO (cutting-surface / exchange methods, dual reformulations).

Key contributions

  • 2019 / 2022 — DRO survey (with Rahimian). The 90+ page Open Journal of Mathematical Optimization review. The principal reference for the wiki's DRO concept page.
  • 2018 — DRO for statistical learning (with Zhang). One of the first papers to show that distributionally robust classification / regression over a Wasserstein ambiguity set is equivalent to a regularised ERM problem; connects to SVM, LASSO, regularised logistic regression. Cited in the survey's §3.4 and §5.1.1.3.
  • Cutting-surface / exchange methods for semi-infinite programs. Co-author of foundational SIP algorithms (with Papp) that underpin the practical solving of DRO inner-sup problems.
  • DRO in two-stage stochastic programs. Long-standing line of work on robust counterparts of two-stage stochastic linear programs with recourse.

Why this matters for the wiki

Mehrotra's work is the theoretical scaffolding behind any deployable DRO-based risk model. The wiki's DRO page, Wasserstein page, and coherent-risk page are all grounded in his survey. His work on SIP exchange methods is the bridge between the abstract DRO formulation and the computational tractability required for production risk models.

See also