You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@commons.apache.org by "Luc Maisonobe (JIRA)" <ji...@apache.org> on 2011/07/29 17:46:10 UTC
[jira] [Resolved] (MATH-635) High order Brent-like bracketing
solver
[ https://issues.apache.org/jira/browse/MATH-635?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Luc Maisonobe resolved MATH-635.
--------------------------------
Resolution: Fixed
Fixed in subversion repository as of r1152276.
> High order Brent-like bracketing solver
> ---------------------------------------
>
> Key: MATH-635
> URL: https://issues.apache.org/jira/browse/MATH-635
> Project: Commons Math
> Issue Type: New Feature
> Reporter: Luc Maisonobe
> Assignee: Luc Maisonobe
> Fix For: 3.0
>
>
> The new bracketing solvers greatly improve usage. Unfortunately, for now there are only secant-based bracketing solvers available, and a wrapper which basically ends by adding a few secant steps to regular non-bracketing solvers.
> Changing the Brent solver to provide bracket selection on the final result would depart from the standard algorithm, which is a wrong move (see MATH-599 for a similar case recently resolved).
> It would be nice to set up another solver in the same spirit as Brent (i.e. using inverse polynomial interpolation when possible and falling back to dichotomy) while retaining bracketing. A nice and simple improvement is also to use higher order inverse polynomial interpolation by retaining several previous points. This allows to build a solver that have an higher order than Newton for example but still only needs function values and no derivatives at all.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira