Implementing a new benchmark (product quantization)

[unmeshvrije]

There is a paper about approximate nearest neighbours which can be found here

Previously, we implemented this algorithm in C++, however since we have all our codebase in Python, we would like to use the Python implementation for this algorithm.

Fortunately, there is already one pure python implementation for the same.

In this issue, we should use this library/module in our code and compare the performance of our system with approximate nearest neighbours.

e.g. For a query h, r, ? our method (subgraphs) would produce certain top K results, and we would like to compare these results with the nearest neighbours of the vector h op r where op is the operation defined by the embedding model