oertl / hyperloglog-sketch-estimation-paper Goto Github PK

Hi there!

I read through your wonderful paper, which I discovered because of the reference to it in the Redis source code (of course, you know this because you wrote it in redis/redis@1e9a774#diff-d2bbaa446905d084e2b9d0bb10b299bb87a465a7cdfd95768aa2d320db357dc2). I am surprised that your improved estimator is not more well known among sketching enthusiasts. It seems after Kane-Nelson-Woodruff, and the Heule-Nunkesser-Hall paper from Google, academics are not paying as much attention to the practical side.

I would like to cite your work in a book I'm writing, which will include a small section on HyperLogLog. Do you have a preferred citation? I was not able to find a conference or journal that this was published in. Is it accepted for publication in a future conference or journal, by chance? If not, I can refer to arXiv. It would be as follows:

@Article{Ertl17,
  author  = {Ertl, Otmar},
  journal = {arXiv preprint arXiv:1706.07290},
  title   = {New Cardinality Estimation Methods for HyperLogLog Sketches},
  year    = {2017},
}

And please let me know if there's anything specific you'd like me to include related to your work.

Hi:
what would be the license for your code?
Thanks!

I have unsuccessfully attempted to build the paper and a variety of the sub tasks in the build.gradle file. Can you provide dependency and build information so that others can replicate the work here?

I find it non trivial to compare figures between multiple estimation functions. I would recommend to take a look at multiple boxplot for achieving this, e.g. https://pvk.ca/Blog/2012/07/03/binary-search-star-eliminates-star-branch-mispredictions/

There's also this variant: https://pvk.ca/Blog/2014/04/13/k-ary-heapsort/