commons-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Bruce A Johnson (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (MATH-1092) NonLinearConjugateGradientOptimizer's Line search is a gradient search returns obviously suboptimal point.
Date Sun, 02 Mar 2014 21:12:31 GMT

    [ https://issues.apache.org/jira/browse/MATH-1092?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13917559#comment-13917559
] 

Bruce A Johnson commented on MATH-1092:
---------------------------------------

Couldn't we have another constructor that takes a fourth argument (the initial upper limit
for the bracket search).  The current constructor could default to 1.0 for that value so existing
tests would pass.

The problem I'm working on is the energy minimization of a complex molecule.  I'm sure there
are many local minima and if the bracket is too large it finds a minimum that is not as good
as one near the origin of the search.  So it seems essential to be able to specify the size
of the search region. (As an aside, I use the CMAESOptimizer to initially search in this complex
landscape, but (when I can use gradients) the ConjugateGradientOptimizer is useful to quickly
refine the solution. 

> NonLinearConjugateGradientOptimizer's Line search is a gradient search returns obviously
suboptimal point.
> ----------------------------------------------------------------------------------------------------------
>
>                 Key: MATH-1092
>                 URL: https://issues.apache.org/jira/browse/MATH-1092
>             Project: Commons Math
>          Issue Type: Bug
>            Reporter: Ajo Fod
>         Attachments: MATH-1092.patch
>
>
> In package : org.apache.commons.math3.optim.nonlinear.scalar.gradient
> In a minimization problem, a line search should not return a point where the value is
greater than the values at the edges of the interval. The line search violates this obvious
requirement by focusing solely on solving for gradient=0 and ignoring the value.
> Moreover LineSearchFunction is something that can be used in other contexts, so perhaps
this should be a standalone class.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message