Andrea Lodi, the Andrew H. and Ann R. Tisch Professor at Jacobs Technion—Cornell Institute at Cornell Tech and a member of the School of Operations Research and Information Engineering graduate field... Read more about ORIE’s Lodi named winner of prestigious Farkas Prize
DOD awards ORIE’s Gunluk grant to study interpretable binary classification
Discrete optimization techniques have been successfully applied to many industrial applications such as production planning, scheduling and logistics. However, these techniques—especially integer linear programming (IP)—are not yet widely used in AI applications even for those that are combinatorial in nature (e.g., classification or clustering). An important reason behind this is the presence of large data sets in AI applications, which usually lead to large-scale IPs that are hard to solve.
ORIE Professor of Practice Oktay Gunluk has received a grant of more than $450,000 from the U.S. Department of Defense/Office of Naval Research to study integer programming models for interpretable binary classification.
“With the use of machine learning (ML) techniques to automate socially and legally sensitive decision-making tasks such as lending, hiring, and college admissions, the need for interpretable ML models is increasing” Gunluk said. “Moreover, in most of these applications, the classification algorithms are required to be fair to all those affected.”
As one would expect, there is usually a trade-off between interpretability/fairness and predictive accuracy. This trade-off can be modelled using an IP formulation, which has an exponential size. Gunluk and Ph.D. student Connor Lawless would study computational techniques to solve these large-scale formulations and investigate related theoretical problems. They would also collaborate with researchers from the IBM T. J. Watson Research Center and Wisconsin University.