matlab - How to bucket locality-sensitive hashes? -

i have algorithm produce locality-sensitive hashes, how should bucket them take advantage of characteristics(i.e. similar elements have near hashes(with hamming distance))?

in matlab code found create distance matrix between hashes of points search , hashes of points in database, simplify code,while referencing called charikar method implementation of search method.

i tried search that, i'm not sure how apply case of methods found(like multi-probe method). none of these techniques seems pluggable if have hashes. there simple example code this? or suggestion?

this link page matlab code i'm talking about: http://www.eecs.berkeley.edu/~kulis/klsh/klsh.htm

based on: search in locality sensitive hashing this, after reading similarity estimation techniques rounding algorithms:

this question somehow broad, going give minimal (abstract) example here:

we have 6 (= n) vectors in our dataset, d bits each. let's assume 2 (= n) random permutation.

let 1st random permutation begin! remember permute the bits, not order of vectors. after permuting bits, maintain order, example:

v1 v5 v0 v3 v2 v4 

now query vector, q, arrives, it's (almost) unlikely going same vector in our dataset (after permutation), won't find performing binary search.

however, going end between 2 vectors. can imagine scenario (for example q lies between v0 , v3:

v1 v5 v0 <-- pointer    <-- q lies here v3 <-- down pointer v2 v4 

now move either or down pointer, seeking vi vector match @ bits q. let's v0.

similarly, second permutation , find vector vi, let's v4. compare v0 first permutation , v4, see 1 closest q, i.e. 1 has bits equal q.

---

however, if seeking ready implementation, should ask in software recommendation. @ paper linked to see if author(s) made code public, or if share after contacting them.