Search
06/01/2015 - 14:30 to 14:50
Stage 1
short talk (20 min)
Intermediate
Session abstract:
The t-digest is a state-of-the-art algorithm for computing approximate quantiles with adjustable accuracy limits and very few limitations.
Implementations of t-digest algorithm are easy to use and have been integrated in all kinds of software from ElasticSearch to Apache Mahout. Certain kinds of queries such as finding the top 99.999th %-ile can be accelerated by several orders of magnitude by using t-digest.
I will describe the basic algorithm and demonstrate the effect of some variations of the algorithm. I will also show how to use the algorithm in your code or your queries.