Risk, Recourse, and Resilience Modeling via Stochastic Combinatorial Optimization

DANIEL J. EPSTEIN DEPARTMENT OF INDUSTRIAL AND SYSTEMS ENGINEERING SEMINARGuest Speaker: Dr. Suvrajeet SenProfessor of ISE, Ohio State UniversityABSTRACT: The stochastic programming literature has been mainly devoted to continuous stochastic optimization models in which discrete decisions are not allowed. However, there are numerous applications, especially those arising in the context of decision-making for homeland security problems, in which discrete choices are natural. In this talk, I will present several applications for which the use of stochastic combinatorial optimization (SCO) is absolutely essential. Yet, it is only within the last 10 years during which researchers have begun to investigate methods for solving SCO problems without brute-force enumeration. In this lecture, I will begin by discussing three applications. One application arises from the need to model the Kahneman/Tversky "S-curve". Another application deals with a stochastic "server" location problem in which servers are to located before demand seems firm. And a final application will be devoted to network design under threat. These applications lead to models that are so large, that even the most powerful MIP solvers are unable to provide a feasible solution. After motivating the need for new methodology to handle such problems, we will discuss some algorithmic strategies that are the only ones to be able to solve these very large scale optimization models. On the other hand, advances in decomposition-coordination methods have made it possible to solve these very large scale models. It is worth noting that the most well known deterministic solvers are unable to solve most of the challenging instances. The keys insights leading to the SCO decomposition rely on effective use of successive convexification. We will summarize a collection of algorithms that have been developed for the solution of such problems, with the largest of these containing over a million binary variables! MONDAY, MAY 11, 2009, ELECTRICAL ENGINEERING BLDG (EEB) ROOM 248, 10:30 – 11:30 AM

Location: Hughes Aircraft Electrical Engineering Center (EEB) - 248

Audiences: Everyone Is Invited

Contact: Georgia Lum