benwtrent commented on PR #14078:
URL: https://github.com/apache/lucene/pull/14078#issuecomment-2679008833

   @gaoj0017 
   
   > The OSQ method (introduced in this PR) has its major idea similar to our 
extended RaBitQ method and our extended RaBitQ method is a prior art which 
achieves good accuracy at 1-bit/2-bit binary/scalar quantization for the first 
time.
   
   Thank you for further clarifying what y’all are meaning by prior art, this 
helps us understand each of our “sticking” points.
   
   Extended RaBitQ is, to our knowledge, the first thing to do something other 
than min/max quantization per vector. Indeed, extended RaBitQ has impressive 
numbers from y’all’s paper and work. 
   
   Our sticking point is that the quantization optimizations for OSQ are 
unique. While both are improving the quantization quality per vector, OSQ is 
going about it in a different way. Its direct scalar quantile optimization via 
anisotropic loss. Both of which find their roots in the linked research in the 
PR.
   
   I will happily indicate in the code comments that extended RaBitQ is the 
first to do this more advanced quantization per vector. Because to our 
knowledge it is true.
   
   We want to donate this to the Apache Lucene project and community so it can 
grow and evolve here.
   
   Additionally, we are wanting to do a broader survey work (though likely not 
an academic article), that shows the strengths and improvements over all of the 
major quantization techniques, including extended RaBitQ, LVQ, PQ, and OSQ. I 
think a direct comparison between the techniques helps provide color for the 
strengths of each.
   


-- 
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

Reply via email to