algorithm - locality sensitive hashing for infinite feature space -
i'm trying wrap head around locality-senstive hashing in case when can not enumerate possible features (e.g. facebook likes when comparing users). there solutions adressing problem?
the locality-sensitive hashing algorithms i've seen far depend on finite vectors of length k
k
total number of features (e.g. words). in case, not know total set of features beforehand, still want find n-nearest neighbours new item in database. given targeted size of database, re-calculating pairwise similarity each insert not feasible.
how can tackle problem? has encountered similar problem , found solution?
Comments
Post a Comment