Optimal Las Vegas Approximate Near Neighbors in lp
Speaker: AlexTitle: Optimal Las Vegas Approximate Near Neighbors in lp
Date: 29 Apr 2019 5:30pm-7:00pm
Location: Maxwell-Dworkin 221
Food: Indian
Audience background: General probability background helps, but I plan to give all necessary definitions
Abstract: Nearest neighbor search is the algorithmic problem where, given a database D of points, we want to support queries that ask for the point in D closest to a query point. In order to avoid the “curse of dimensionality” present for (exact) nearest neighbor search, one can formulate an approximate version of the nearest neighbor search problem. There has been a rich line of work over the past two decades developing efficient solutions to approximate near neighbor search. In this talk, I will give an overview of two techniques for this problem—locality-sensitive hashing and product quantization (a.k.a. tensoring)—and show how they come together to construct optimal Las Vegas algorithms for approximate near neighbor search. (Will feature some fun applications of dimensionality reduction!)