10/19/2017 | 4:15pm | E51-335

Reception to follow.

Mohit Tawarmalani

Professor
Krannert School of Management, Purdue University

Abstract: We develop a new relaxation that exploits function structure while convexifying a product of n functions. The function structure is encapsulated using at most d over and underestimators. We convexify the function product in the space of estimators. The separation procedure generates facet-defining inequalities in time polynomial in d for a fixed n. If the functions are non-negative, the concave envelope can be separated in O(n d log(d)). Then, we extend our construction to infinite families of under and overestimators. Our relaxation procedure can be interpreted as a two-step procedure where we first express the product as a telescoping sum and in the second step apply a simple relaxation strategy. This interpretation admits various generalizations that yield various valid inequalities for nonconvex programs. We conclude by discussing techniques to generate the over and underestimators and various ways in which the proposed techniques improve and/or generalize current relaxation schemes for factorable programs. 

Bio: Mohit Tawarmalani is a Professor and Allison and Nancy Schleicher Chair at the Krannert School of Management, Purdue University. He received his PhD and master's from University of Illinois at Urbana-Champaign and bachelor's from Indian Institute of Technology, Delhi. Mohit has worked in MIS at Tata Engineering and Locomotive company and in geometric modeling at ComputerVision. He teaches courses in business analytics and optimization and is the current academic director of Masters in Business Analytics and Information Management at Purdue University.

Mohit Tawarmalani's research interests are at the interface of computer science, optimization, and their applications in business and engineering. Mohit has co-authored a book on global optimization algorithms and has co-authored BARON, a software for global optimization. Mohit was awarded the 2004 INFORMS Computing Society Prize, the 2005 Best Paper Award from the Workshop on Information Technology and Systems, and the 2006 Beale-Orchard-Hays Prize from the Mathematical Programming Society. He is also the 2003 Jay N. Ross Young Faculty Scholar and the 2006 Krannert Faculty Fellow.

Mohit serves as an Associate Editor for the Journal of Global Optimization and as a Technical Editor for Mathematical Programming Computation.