Abstract
Topic: Bridging Optimization and AI: Some recent advances
The confluence of learning and optimization holds great promise for solving dynamic, online resource allocation problems under uncertainty. In this talk, I will present an overview of two such problems in which tools developed independently in one field have helped solve problems in the complementary field.
First, we focus on the problem of collaborative routing of aircraft and unmanned vehicles, under non-stationary conditions with spatial-temporal correlations. A long posited question in collaborative routing has been: “if a subset of traveling vehicles can be used for exploring parts of the network where information has become sparse, how should vehicles be routed to collect information most useful to minimize costs for the entire fleet”? We expand on recent advances in multi-armed bandit algorithms to solve this problem, and generate near-optimal policies for schedule-dependent and schedule-independent exploration. We find this could reduce travel time and fuel burn by about 5% for the fleet.
Second, we study how we can improve upon long-used regret bounds in AI/ML, which are based on omniscient information about the future. Our work builds on Brown, Smith, and Sun (2010), who present an approach for information-relaxation bounds in dynamic programs. Their approach penalize the additional information available to the anticipative (omniscient) decision-maker who knows the future. However, the computation of these penalized information-relaxation bounds (or penalized regret bounds) for large-scale systems has hitherto been discussed by many as intractable. We present tractable methods to achieve bounds for time-space network-based problems and problems modeled as block-diagonal mixed-integer programs.
Bio
Lavanya Marla is an Associate Professor in Industrial and Enterprise Systems Engineering at the University of Illinois at Urbana-Champaign. Her research interests are in robust and dynamic decision-making for large-scale networks subject to operating stochasticity. She builds cross-cutting methodologies that integrate data-driven optimization, statistics, simulation and machine learning. Application areas of interest include aviation planning, operations and pricing; logistics, emergency medical services, and shared transportation systems. Prior to the University of Illinois, she was with the Heinz College at Carnegie Mellon University. She earned her PhD in Transportation Systems, and dual Masters in Operations Research and Transportation from the Massachusetts Institute of Technology; and previously, a Bachelors degree from IIT Madras. Her work has been recognized through the prestigious Center for Advanced Study award from the University of Illinois, as a semi-finalist at the INFORMS Innovative Applications in Analytics Award, research awards from the International Conference for Research in Air Transportation, AGIFORS, Knowledge Discovery and Data Mining (KDD) and others. Her work is funded by the National Science Foundation, the Department of Homeland Security, the Department of Transportation, the US-India Educational Foundation's and multiple industry grants. She has served in multiple leadership roles with the INFORMS Transportation Science and Logistics Society, Women in OR/MS and AGIFORS.