msokolov commented on PR #947: URL: https://github.com/apache/lucene/pull/947#issuecomment-1180963745
I'm looking to address various comments; just pushed a commit that makes the vector encoding explicit by adding a new enum and parameter "vectorEncoding", splitting this out from "similarityFunction". > During merging when writing a merged vector field it looks like we first expand vector values only to again to compress them later? Would be nice to avoid this. Oh good catch, @mayya-sharipova I will look into addressing this. > Not sure if possible at this stage, but it would be nice if HnswGraphBuilder and HnswGraphSearcher were not aware of different calculations needed for different similarity functions, and refer all this calculations (dotProduct(BytesRef a..)) to VectorSimilarityFunction I don't see how to do this efficiently (without many conversions from byte to float) and neatly (without code duplication in tricky algorithmic areas) and with complete API purity, so I sacrificed some purity. If you have any ideas how to do it better, I'm open to changing it though. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
