# Race Sketches This last one is from the "cutting edge" of computer science research on sketches. It's an algorithm that uses locality-sensitive hash (LSH) functions ([http://tylerneylon.com/a/lsh1/](http://tylerneylon.com/a/lsh1/)) and a bunch of count-min-sketches to build an estimate of kernel density in high dimensional spaces. This is a notoriously computationally intensive problem in stats: https://randorithms.com/2020/09/15/RACE-KDE.html ### Some Details See more in [Locality Sensitive Hashing (LSH)](Locality%20Sensitive%20Hashing%20(LSH).md). ### Questions on implementation * We will want to chose the correct LSH family that is appropriate for the similarity measure we wish to preserve.For example, if you're interested in angular similarity, you might choose a hash function family that is sensitive to the angle between vectors. --- Date: 20210813 Links to: [Computer Science MOC](Computer%20Science%20MOC.md) [Algorithms and Data Structures](Algorithms%20and%20Data%20Structures.md) Tags: #todo References: * [Main overview](https://randorithms.com/2020/09/15/RACE-KDE.html)