Open gingerwizard opened 2 years ago
One way of looking at selection - it doesn't consider all algorithms. Taken from https://towardsdatascience.com/comprehensive-guide-to-approximate-nearest-neighbors-algorithms-8b94f057d6b6
Useful overviews -
https://arxiv.org/abs/1610.02455 - sections 2-6 cover most algorithm options
https://arxiv.org/abs/1612.07545 - focus on hashing options
Interesting analysis https://medium.com/vespa/approximate-nearest-neighbor-search-in-vespa-part-1-36be5f43b8d1. Vespa is probably the leading vector search product out there.
This is super useful! https://github.com/facebookresearch/faiss/wiki/Guidelines-to-choose-an-index
ANN algorithms benchmarking: http://ann-benchmarks.com/
This attempts to consider some of the criteria we should consider when selecting algorithm:
We should in turn state our requirements, based on the above, and filter to those algorithms which satisfy them.