Indexing dense numeric vector

[huahaiy]

Add a data type :vec, for indexing dense numeric vectors and search based on similarity.

[huahaiy]

Current state of art is ScaNN, see paper, extended paper, which is built for tensorflow serving and depends on tensorflow.

Prior literature:
https://lear.inrialpes.fr/pubs/2011/JDS11/jegou_searching_with_quantization.pdf
http://proceedings.mlr.press/v51/guo16a.pdf
https://www.amazon.com/Quantization-Compression-Springer-International-Engineering/dp/0792391810

[huahaiy]

ScaNN is fast for it does lookup table in SIMD register: https://medium.com/@kumon/similarity-search-scann-and-4-bit-pq-ab98766b32bd (paper https://arxiv.org/pdf/1704.07355.pdf)

Faiss has better packaging, also, it has PQ4 implementation as well. https://github.com/facebookresearch/faiss/wiki/Indexing-1M-vectors#4-bit-pq-comparison-with-scann

[huahaiy]

For our initial version, https://github.com/nmslib/hnswlib seems to be ideal, for the following reason:

1. It performs well across the board. https://github.com/erikbern/ann-benchmarks/
2. It out performs everything else for huge vectors https://ann-benchmarks.com/kosarak-jaccard_10_jaccard.html. The current batch of state of the art libraries optimize for vectors less than 1000 dimensions. But my hunch is that huge vectors (greater than 10k dimensions) are needed to really take advantage of the vector views of semantics (i..e. the so called vector symbolic approach).
3. It is small and seems to be much easier to integrate than others.

[huahaiy]

Another option, is https://github.com/unum-cloud/usearch