New ANN libraries much faster than flann!

[fengweichangzi]

I think this project does not update anymore.
I tried to find some other similar library and found one benchmark that can compare all ANN libraries.

Well, flann performs not so good.
https://erikbern.com/2018/02/15/new-benchmarks-for-approximate-nearest-neighbors.html

Maybe it is time for us to try some new ANN libraries.
E.g., HNSW, KGraph, SW-graph,FAISS-IVF Annoy and so on

[stephematician]

That's a great resource.

It's hard to say whether FLANN is slower or faster because it depends on the size of the index, the dimensions, the algorithm used, the number of queries, the precision required, and things like the entropy/mutual information of the data - there are many factors!

In problems where the number of queries >> size of the index, at higher dimensions, HNSW and others are probably faster than FLANN - but it can still vary depending on some of those factors listed.

To make it easier to find - more up-to-date comparisons of ANN can be found via https://github.com/erikbern/ann-benchmarks/.