Ole

@glicerico Do you want to create a pull request for that edit?

@timgates42 If you still have the time for it, could you please create the pull request? Thank you!

@uriva I have no experience with using Hilbert curves in this application. If you are interested in adding a hilbert curve based hash to NearPy feel free to do so...

@VsevolodZakharov Thanks for the hint. I try to find the time.

Not really yet. If the combined candidate list of all buckets that were "hit" by the hashes has less then k vectors, then this is not possible yet. If the...

Alright. We might have to do some experiments regarding speed of NearPy. As far as I remember (have not touched code base since last commit) the sparse representation in the...

So you want to always get k results independent of the query vector? As long as the density of your dataset is almost everywhere the same, this can be reached...

@piskvorky So regarding next steps. Of course help and working together would be great. If you interested in this you could check out papers about density-sensitive LSHs, or "Density Sensitive...

@piskvorky Maybe using a tree structure in each individual LSH would solve that. kd-tree or range tree or something (just some buzzwords. i have to check on those details first)....

My pleasure. I really like this topic and would enjoy it, if NearPy would improve collaboratively.