DominikHorn / learned-hashing

A header only cmake/c++ library that exposes various state-of-the-art learned hash functions, i.e., learned models that are repurposed to fill the role of a classical hash function.

Learned Hashing

A header only cmake/c++ library that exposes various state-of-the-art learned hash functions, i.e., learned models repurposed to fill the role of a classical hash function.

Learned hash functions can for example aim to utilize knowledge about the underlying data distributions to minimize collisions. One common technique for achieving this is to learn the CDF of the dataset and use CDF(x) as the hash function. Since CDF functions are monotone, such a hash function is also automatically monotone:

k1 <= k2 --> h(k1) <= h(k2)