Baunsgaard commented on issue #875:
URL: https://github.com/apache/systemml/pull/875#issuecomment-617192727
> Hi @Baunsgaard , let us move this forward. With each improvement, if we
track the test results in this PR itself in the phase by phase manner (Might
take a few weeks :), like with any good PR.).
>
> So, for this basic setup _shall we track_:
>
> 1. computational cost (how much lower?)
>
> 2. The accuracy of DV estimate!
>
>
> Once we are satisfied with the results, we could request perhaps _Berthold
Reinwald_ for a more involved review. Thanks.
currently this pr is on a hold, because of multiple reasons:
1. The intention was to implement a function and use it in the compression
planning as well, so that we have more granularity in our estimations, and not
relying only on the current compression sample estimators. Therefore i want to
make the compression planning accurate first without estimations before coming
back to these estimation algorithms.
2. The accuracy achieved with the current implementation of KMV so far
achieves worse estimates than the ShlosserJackKnifeestimator inside the
compression.
3. I'm working on HyperLogLog ( and variations) on the side to compare with
this as well.
I see no reason to contact anyone, and thinks it will be counter productive
in this specific case, since the target for these implementations at the moment
is exploratory in nature.
----------------------------------------------------------------
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:
[email protected]
