This document discusses locality-sensitive hashing (LSH) for similarity search in high dimensions. It introduces LSH as a technique to "hash" similar items to the same bucket with high probability. Locality-sensitive functions are defined that map similar items to the same value more often than dissimilar items. The banding technique divides the data into bands to improve performance. LSH families are described for cosine similarity. Applications include near neighbor search, entity resolution, and matching fingerprints or articles.