jmalkin commented on issue #647: URL: https://github.com/apache/datasketches-java/issues/647#issuecomment-2746742919
I tried a basic simulation of the cheaper sampling approach, in this case using octave to track simply the number of times a sample was added to the reservoir. I averaged across many trials (things at 100k didn't seem so different from 1-2M, happily). There seems to be a small but consistent bias using the faster approach of picking the number of items to skip. I used small reservoirs of 5 and 7, and only went out to `n` of a few hundred but the bias seems to be consistent -- and to grow slowly over time. Since I can prove the existing algo's probabilities are correct, I'm treating any deviations from that as suspicious and likely problematic. -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
