Learning-Augmented Algorithms
In recent years there has been increasing interest in using machine learning to improve the performance of classical algorithms in computer science, by fine-tuning their behavior to adapt to the properties of the input distribution. Many applications involve processing streams of data (video, data logs, customer activity etc) by executing the same algorithm on an hourly, daily or weekly basis. These data sets are typically not "random" or "worst-case"; instead, they come from some distribution which does not change rapidly from execution to execution. This makes it possible to design better algorithms tailored to the specific data distribution, trained on past instances of the problem. Using this “data-driven” or “learning-augmented” approach to algorithm design, our group members design better data structures, online algorithms, streaming and sublinear algorithms, algorithms for similarity search and inverse problems.
Faculty: Costis Daskalakis, Piotr Indyk, Ronitt Rubinfeld
Postdocs: Talya Eden, Anders Aamand
Students: Justin Y Chen, Shyam Narayanan, Sandeep Silwal, Nicholas Schiefer
Alumni: Tal Wagner, Ali Vakilian, Ilya Razenshteyn, Yang Yuan
Learning-based Support Estimation in Sublinear Time
T Eden, P Indyk, S Narayanan, R Rubinfeld, S Silwal, T Wagner
International Conference on Learning Representations (ICLR), 2021.
Learning Sublinear-Time Indexing for Nearest Neighbor Search
Y Dong, P Indyk, I Razenshteyn, T Wagner
International Conference on Learning Representations (ICLR), 2020.
Learning-based low-rank approximations
P Indyk, A Vakilian, Y Yuan
Advances in Neural Information Processing Systems (NeurIPS), 2019.
Learning-Based Frequency Estimation Algorithms
CY Hsu, P Indyk, D Katabi, A Vakilian
International Conference on Learning Representations (ICLR), 2019.
Learning-augmented algorithms course lecture notes
https://stellar.mit.edu/S/course/6/sp19/6.890/materials.html
FODSI Workshop on Machine Learning for Algorithms