Open wdongyu opened 1 month ago
We are computing similarity by the inner product, so yes, you can do that for the datasets in this repo.
The billion-scale ann benchmarks project contains additional datasets: https://github.com/harsha-simhadri/big-ann-benchmarks for inner product similarity.
Question
Can I use the "ground truth" part of angular datasets for inner product, after normalizing the "data" part. If not, are there any datasets for inner product? I have found music-100, but it seems unavailable now :(