0.1.0-PRE30

• Added new submodules which can be used without Elasticsearch:
  • com.klibisz.elastiknn:models contains exact and approximate similarity models, all in Java with minimal dependencies.
  • com.klibisz.elastiknn:lucene contains the custom Lucene queries and some Lucene-related utilities used by Elastiknn.
  • com.klibisz.elastiknn:api4s_2.12 contains Scala case classes and codecs representing the Elastiknn JSON API.
  • com.klibisz.elastiknn:client-elastic4s_2.12 contains the Elastiknn Scala client.

0.1.0-PRE29

• Switched to less-naive implementation of multiprobe L2 LSH. Specifically, uses algorithm 1 from Qin, et. al. to generate
  perturbation sets lazily at query time instead of generating them exhaustively. This does not use the estimated
  scoring optimization from that paper.
• Performance optimizations for approximate queries. Specifically using a faster sorting method to sort the hashes before
  retrieving matching docs from the shard.

0.1.0-PRE28

• Introduced multiprobe L2 LSH. It's a small change. Search for probes in the API docs.
• Bug fix for an edge case in approximate queries.

0.1.0-PRE27

• Changed to GPLv3 license. See: https://github.com/alexklibisz/elastiknn/blob/master/LICENSE.txt.

0.1.0-PRE26

• Added Permutation Lsh model and query, based on paper Large Scale Image Retrieval with Elasticsearch by Amato, et. al.
• Several internal improvements, including support for LSH models with repeated hashes.

0.1.0-PRE25

• Performance improvements for LSH queries. 1.5-2x faster on regular benchmarks with randomized data. See PR #114.

0.1.0-PRE24

• Fixed error with KNN queries against vectors that are stored in nested fields, e.g. outer.inner.vec.

0.1.0-PRE23

• Switched LSH parameter names to more canonical equivalents: bands -> L, rows -> k,
  based on the LSH wikipedia article
  and material from Indyk, et. al, e.g. these slides.
• Added a k parameter to Hamming LSH model, which lets you concatenate > 1 bits to form a single hash value.

0.1.0-PRE22

• Switched scala client to store the ID as a doc-value field. This avoids decompressing the document source
  when reading results, which is about 40% faster on benchmarks for both exact and approx. search.

0.1.0-PRE21

• Re-implemented LSH and sparse-indexed queries using an optimized custom Lucene query based on the TermInSetQuery.
  This is 3-5x faster on LSH benchmarks.
• Updated L1, and L2 similarities such that they're bounded in [0,1].