MinHash
A probabilistic method that produces compact signatures for estimating Jaccard similarity between large sets without comparing every element pair directly.
Direct Jaccard comparison becomes expensive when many large document or feature sets must be compared. MinHash stores minimum values under multiple hash permutations; the fraction of matching signature positions estimates the Jaccard similarity of the underlying sets.
It is not a universal replacement for vector-distance search. Weighted features and dense embeddings often need different methods. MinHash is especially useful for shingle/set representations and duplicate or near-duplicate discovery, while FAISS vector similarity search addresses dense-vector retrieval.