search

maumueller / ann-benchmarks-sisap19

http://ann-benchmarks.com/sisap19/
MIT License
2 stars 0 forks source link

Running benchmarking on separately constructed binary indexes. #1

Open princethewinner opened 4 years ago

princethewinner commented 4 years ago

I have binary indexes (generated from an unknown source) for a dataset. I want to find nearest neighbors for some query examples using hamming distances. Is it possible to use FLANN, MIH, HNSW or Annoy from this benchmarking?

maumueller commented 4 years ago

Dear Prashant

Yes, this is possible. You first have to convert your dataset to the hdf5 format in ann-bechmarks. Several examples of how to do this can be seen here: https://github.com/maumueller/ann-benchmarks-sisap19/blob/master/ann_benchmarks/datasets.py#L220 (and below)

You should then be able to run the algorithms on your dataset in the normal way. This is how the bit indexes parameters are specified: https://github.com/maumueller/ann-benchmarks-sisap19/blob/master/algos.yaml#L544