Thread by @leland_mcinnes

Pynndescent, an approximate nearest neighbor search library, got a major update recently. Index construction is now multicore by default. Querying is now much faster -- competitive with some of the fastest ANN libraries around.
(1/4)

Performance in particularly strong for higher accuracy (>90%) queries.
(2/4)

The library also comes equipped with a Transformer class fitting in with the new KNeighborsTransformer in scikit-learn (scikit-learn.org/stable/modules…) to allow you to speed up various sklearn models and pipelines.
(3/4)

A big thank you to all the contributors (particularly @tom_e_white), and to @fulhack and Martin Aumüller for their excellent ANN-benchmarking suite (github.com/erikbern/ann-b…)
(4/4)

@tom_e_white @fulhack pip install pynndescent

conda install -c conda-forge pynndescent

github.com/lmcinnes/pynnd…

(5/4)

Also a big thank you to the @numba_jit team that made this performance possible. The recent addition of TypedLists was particularly valuable. Pynndescent would not be the same without it!
(6/4)