When you live in the big city, finding everyone’s nearest neighbours takes a long time! Our judicious city planners have conjured a technique using randomness, hyperplanes, buckets, lexical sorting, and binary numbers to get the job done approximately and fast. Really fast. They call it locality-sensitive hashing and management loves the idea. But it sounds too good to be true!
In this talk we’ll explore locality-sensitive hashing, a technique to turn the computationally expensive exact nearest-neighbor search problem into an inexpensive approximate solution (it’s a neat trick and I promise you’ll love it). We’ll see how locality-sensitive hashing is used in image search, recommendations, and other machine learning problems. And of course, we’ll mention deep hashing, because why not?
Aaron Levin is a mathematician who fell in love with programming and now manages Data Science teams at SoundCloud. Aaron found a rare record once and wrote about it on Weird Canada.