commons-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ajo Fod (JIRA)" <>
Subject [jira] [Commented] (MATH-418) add a storeless version of Percentile
Date Mon, 17 Jun 2013 17:03:20 GMT


Ajo Fod commented on MATH-418:

Any solution is better than the current situation. However, there are two desirable features.

Reducability: Since large data datasets typically are associated with parallel environments,
the ideal algorithm would be map/reduce-able. 

Heteroskedasticity: What if one passes it the quantiles of say an exponential distribution
not randomly but in sequential order to the quantile estimator? How big would the quantile
error be? Is it significantly more than the iid case?
> add a storeless version of Percentile
> -------------------------------------
>                 Key: MATH-418
>                 URL:
>             Project: Commons Math
>          Issue Type: New Feature
>    Affects Versions: 2.1
>            Reporter: Luc Maisonobe
>             Fix For: 4.0
> The Percentile class can handle only in-memory data.
> It would be interesting to use an on-line algorithm to estimate quantiles as a storeless
> An example of such an algorithm is the exponentially weighted stochastic approximation
 described in a 2000 paper by Fei Chen ,  Diane Lambert  and José C. Pinheiro "Incremental
Quantile Estimation for Massive Tracking" which can be retrieved from CiteSeerX at [].

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see:

View raw message