Re:Show HN: A Go implementation of TensorFlow's streaming qu...
by tsenart on Wednesday August 15, 2018 @06:12PM
(#17770201)
Could you elaborate on the differences a bit deeper? I’m really interested in understanding.
Re:Show HN: A Go implementation of TensorFlow's streaming qu...
by seiflotfy on Wednesday August 15, 2018 @06:25PM
(#17770279)
http://web.cs.ucla.edu/~weiwang/paper/SSDBM07_2.pdf is the paper its mostly based on
Figure 1. Actually describes how big the datastructure can get. It keeps getting bigger the more data you feed it.
T-digest seeks to have a bounded size and an error proportional to q*(1-q), hence it gives up quantile accuracy in the middle of the distribution when under load. This algorithm seems to provide total bounded error without small but unbounded size.