commons-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Gilles (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (MATH-1092) NonLinearConjugateGradientOptimizer's Line search is a gradient search returns obviously suboptimal point.
Date Thu, 27 Feb 2014 15:39:19 GMT

     [ https://issues.apache.org/jira/browse/MATH-1092?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Gilles updated MATH-1092:
-------------------------

    Attachment: MATH-1092.patch

Here is a proposal to factor out the "LineSearch" defined in "PowellOptimizer"
(which uses "BrentOptimizer").
When replacing what needed be in "NonlinearConjugateGradientOptimizer", a series of unit test
started to fail.
I've performed some checks that the convergence criterion is indeed met even though the result
found is now farther from the expected optimum.

The change also uncovered that by using the root solver, the optimizer could not correctly
count the evaluations (cf. "MultiStartMultivariateOptimizerTest").


> 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)

Mime
View raw message