Leland McInnes

Thanks for the reply. My concern was with the input data format for the Jaccard benchmark in ann-benchmarks. It is a (74962 x 27983) sparse matrix. This can be formatted...

Thanks. My initial efforts with the dense matrix are mostly just timing out on the runs, so I fear it is too expensive. Hopefully the sparse jaccard version is not...

Ahead of time compilation creates some packaging issues. Perhaps it could be done, but it is beyond what I can manage right now (I would welcome any help). I have...

It looks like very aggressive application of ``cache=True`` to all the top level numba jitted functions in the various submodules does a good job of alleviating a lot of this....

I suspect you simply need a newer version of numba. It is possible that the requirements.txt and setup.py supply a slightly too old minimum version.

Off the top of my head I don't see any obvious ways to improve this significantly. There is just a lot of computation work to be done, and I'm not...

I'm afraid that at the moment you really need a 2D vector(ish) representation of data for this library to work. I do have some plans to hopefully one day extend...

It is slimly possible that you could manage to do this -- the catch being that the distance computation must be numba compilable. In theory you could write a distance...

My best guess is that it may be due to lack of Python 3.9 support in numba. The latest version 0.53 does support Python 3.9, so you can potentially install...

Thanks for the suggestion. I looked through the ScaNN paper when the preprint first came out, and also at the software they released. It certainly provides very impressive performance (at...