Arpit Agarwal is an Assistant Professor at the CSE Department at IIT Bombay. His research lies in the area of machine learning (ML) and artificial intelligence (AI). His focus is on human-centered AI which includes learning from human feedback, understanding AI impact on individuals and society, and designing socially responsible AI. Prior to joining IIT Bombay, he was a researcher at FAIR Labs (Meta). Before that he was a postdoctoral fellow at the Data Science Institute at Columbia University. He completed his PhD from the CIS Department at University of Pennsylvania, and masters from CSA Department at IISc Bangalore where he was the recipient of the Computer Society of India Medal.

Abstract: Stochastic optimization is a widely used approach for optimization under uncertainty, where uncertain input parameters are modeled by random variables. Exact or approximation algorithms have been obtained for several fundamental problems in this area. However, a significant limitation of this approach is that it requires full knowledge of the underlying probability distributions. Can we still get good (approximation) algorithms if these distributions are unknown, and the algorithm needs to learn them through repeated interactions? Particularly, can we design an online learning algorithm whose performance smoothly approaches the performance of an (offline) optimal algorithm which has knowledge of these distributions?

In this work, we resolve this question for a large class of “monotone” stochastic problems, by providing a generic online learning algorithm with \\sqrt{T log T} regret relative to the best approximation algorithm (under known distributions). Importantly, our online algorithm works in a semi-bandit setting, where in each period, the algorithm only observes samples from the r.v.s that were actually probed. Our framework applies to several fundamental problems in stochastic optimization such as prophet inequality, Pandora’s box, stochastic knapsack, stochastic matchings and stochastic submodular optimization.

In this talk, I will first introduce the multi-armed bandits framework and describe the popular algorithmic principle of "optimism in the face of uncertainty". I will then give a brief introduction to stochastic optimization. Finally, I will describe our result on conversion from offline to online semi-bandit algorithms for stochastic optimization.

This talk is based on a FOCS'24 paper with Rohan Ghuge and Viswanath Nagarajan- https://arxiv.org/pdf/2312.15427.

Date: Tuesday, February 04, 2025
Time: 1:30 PM IST
Venue: AC-02-LR-206-207
For details: ashoka-cdlds@ashoka.edu.in or call: +91-9136857558
Zoom link: https://zoom.us/j/93578528184?pwd=pHITcwwcPOvMLCw44M6WzWf4D7Zj3G.1