Nearest neighbors search can be optimized for specific situations and machines. It would be a good idea to allow the user to specify a mean matching function if they don't want to rely on the default sklearn NearestNeighbors implementation. Over several million candidates, faiss is much faster.
Nearest neighbors search can be optimized for specific situations and machines. It would be a good idea to allow the user to specify a mean matching function if they don't want to rely on the default sklearn NearestNeighbors implementation. Over several million candidates, faiss is much faster.