Instructions for reproducing the GloVe dataset as used in the experiments.
--------------------------------------------------------------------------

We used the GloVe dataset with dimensionality 100 also used in the ann-benchmark: https://github.com/erikbern/ann-benchmarks

1. Download the code from https://github.com/erikbern/ann-benchmarks
2. Go to folder install 
3. execute the script glove.sh 
This will download the original GloVe dataset at your computer.

To reproduce the split of query and probe vectors that we used in our experiments, 
permute the vectors of the original dataset according to the permutation vector perm.txt 
and from the resulting dataset keep the first 100000 vectors as queries and use the remaining as probe vectors.

