[
https://issues.apache.org/jira/browse/MATH-1092?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13914763#comment-13914763
]
Bruce A Johnson commented on MATH-1092:
---------------------------------------
There is a good discussion in Numerical Recipes on using gradients in a line
minimization (the conclusion being basically to use them very judiciously, for
example to figure out what side of the interval to search in). Since we have
gradients available there might be some advantage in optionally using them in
the line minimization). For now, I'm quite happy that with your fix the line
minimizer should never return a point with a higher value than the starting
point (assuming we're doing minimization rather than maximization).
> 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
(v6.1.5#6160)
