You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@datasketches.apache.org by GitBox <gi...@apache.org> on 2020/07/29 07:37:26 UTC
[GitHub] [incubator-datasketches-java] Cheappie opened a new issue #328: Maximal estimate error
Cheappie opened a new issue #328:
URL: https://github.com/apache/incubator-datasketches-java/issues/328
Hi, is there a formula for cpc or hll sketches to calculate max error for estimate ? I wonder what are practices for assuring correctness(high accuracy) of execution while using sketches ?
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@datasketches.apache.org
For additional commands, e-mail: commits-help@datasketches.apache.org
[GitHub] [incubator-datasketches-java] jmalkin commented on issue #328: Maximal estimate error
Posted by GitBox <gi...@apache.org>.
jmalkin commented on issue #328:
URL: https://github.com/apache/incubator-datasketches-java/issues/328#issuecomment-665128666
Error is probabilistic, and for larger values is approximately Gaussian. Bounds estimates are provided in terms of 1, 2, and 3 standard deviations, with a decreasing probability of exceeding each of the bounds, respectively. That same trend persists indefinitely, but we don't quantify it.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@datasketches.apache.org
For additional commands, e-mail: commits-help@datasketches.apache.org
[GitHub] [incubator-datasketches-java] Cheappie commented on issue #328: Maximal estimate error
Posted by GitBox <gi...@apache.org>.
Cheappie commented on issue #328:
URL: https://github.com/apache/incubator-datasketches-java/issues/328#issuecomment-666559339
Ok, now I got It. I was wondering whether I have missed certain information, or It is the way it works. Thank you guys.
@AlexanderSaydakov that's a good point about documentation, for sure the other two sketches could benefit from having documentation as good as theta sketch.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@datasketches.apache.org
For additional commands, e-mail: commits-help@datasketches.apache.org
[GitHub] [incubator-datasketches-java] leerho closed issue #328: Maximal estimate error
Posted by GitBox <gi...@apache.org>.
leerho closed issue #328:
URL: https://github.com/apache/incubator-datasketches-java/issues/328
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@datasketches.apache.org
For additional commands, e-mail: commits-help@datasketches.apache.org
[GitHub] [incubator-datasketches-java] AlexanderSaydakov commented on issue #328: Maximal estimate error
Posted by GitBox <gi...@apache.org>.
AlexanderSaydakov commented on issue #328:
URL: https://github.com/apache/incubator-datasketches-java/issues/328#issuecomment-665132552
As Jon said, for these algorithms there are no hard bounds. There is a distribution of error similar to normal distribution with ever decreasing probability to have an estimate farther and farther away from the true value. The sketch can give you upper and lower bounds for three confidence intervals that roughly correspond to 1, 2 and 3 standard deviations of the normal distribution: roughly 67%, 95% and 99% confidence. There are formulas in the papers. Perhaps we need better documentation on the web site. We have an accuracy table for Theta sketch, but not for HLL and CPC yet.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@datasketches.apache.org
For additional commands, e-mail: commits-help@datasketches.apache.org