@avibryant could do the naive thing and weighted average by event rate of the two streams. Need proofs of error bounds.
-
-
-
@posco intuitively, it feels like that would converge to the mean, not the median. (Imagine doing that for n chunks of 1 event...)
End of conversation
New conversation -
-
-
@avibryant@posco the same as the gradient descent approach to quantiles. Or am I missing something? -
@fdaapproved@avibryant@posco Yeah Robbins/Monro stochastic optimization approach for quantiles is well known. Wish I had this paper... -
@jaykreps@avibryant@posco 2013: the 50s at scale. -
@fdaapproved@jaykreps@posco don't knock it. See also: when do we catch up with 70s programming environments? -
@avibryant@jaykreps@posco somewhere around 2060 at this rate.
End of conversation
New conversation -
-
-
RT
@avibryant: The Frugal 1U streaming quantile estimator is incredibly neat and simple.http://blog.aggregateknowledge.com/2013/09/16/sketch-of-the-day-frugal-streaming/ … -
@samklr@avibryant The AK blog is amazing. Loved the article/demo on HyperLogLog et al. -
@abolibibelot it's my goto place for all things sketches.
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.