Abstract: Traditional online algorithms encapsulate decision making under uncertainty, and give ways to hedge against all possible future events, while guaranteeing a nearly optimal solution as compared to an offline optimum. On the other hand, machine learning algorithms are in the business of extrapolating patterns found in the data to predict the future, and usually come with strong guarantees on the expected generalization error.
In this talk, we will discuss a framework for augmenting online algorithms with a machine learned oracle to achieve competitive ratios that provably improve upon unconditional worst case lower bounds when the oracle has low error. Our approach treats the oracle as a complete black box, and is not dependent on its inner workings or the exact distribution of its errors. We apply this framework to the traditional caching problem -creating an eviction strategy for a cache of size k. We demonstrate that naively following the oracle's recommendations may lead to very poor performance, even when the average error is quite low. Instead we show how to modify the Marker algorithm to take into account the oracle's predictions, and prove that this combined approach achieves a competitive ratio that both (i) decreases as the oracle's error decreases and (ii) is always capped by O(log k), which can be achieved without any oracle input.
This talk is based on this paper, which is joint work with Sergei Vassilvitskii. A preliminary version appeared at ICML'18 and the journal version appeared in the Journal of the ACM.
Bio: Thodoris Lykouris is an assistant professor at the MIT Sloan School of Management where he is affiliated with the Operations Management group and the Operations Research Center. His research focuses on data-driven sequential decision-making and spans across the areas of machine learning, dynamic optimization, and economics. Before joining MIT, Thodoris spent two years as a postdoctoral researcher at Microsoft Research NYC and completed his Ph.D. in 2019 from Cornell University where he was advised by Eva Tardos. His dissertation was selected as a finalist in the Dantzig dissertation award competition. He was also a finalist in the INFORMS Nicholson and Applied Probability Society best student paper competitions. Thodoris is the recipient of a Google Ph.D. Fellowship and a Cornell University Fellowship.
The Data Science Deep Dive is free and open to the public. If you are interested in giving a talk, please email Mohammad Hajiesmaili or Adam Lechowicz. Note that in addition to being a public lecture series, the Data Science Deep Dive is also a seminar (CompSci 692K, Algorithms with Predictions Seminar) that can be taken for credit.