This repository implements a variety of locality sensitive hash functions in Tensorflow 2.0.
The hash functions are implemented in a single module, lsh_functions.py.
We currenly support LSH in several important metric spaces, including
the Euclidean, Manhattan and Lp norm spaces (PStableHash), integer and
string token spaces (IntMinHash and StringMinHash), and inner product
spaces (SRPHash).
You can access these classes by simply importing the LSH function module.
The only dependencies are Python3 and Tensorflow 2.0. The tests use numpy, but this is not required to use the module.
To run the tests, make sure you have Tensorflow and Numpy installed and run:
python3 lsh_functions_test.py
If you find a bug, feel free to create an issue. Ideally, this will include your Python version, Tensorflow version and a self-contained, minimal program to reproduce the bug.
If you would like to request a new LSH function or other functionality, create an issue (or create a pull request with the desired functionality).
This repository is free for research and commercial use, under the Apache-2 license. However, if you find this repo useful, please give us a star or - even better - get in touch!