commons-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Nigel Goodwin (JIRA)" <j...@apache.org>
Subject [jira] [Comment Edited] (MATH-1137) BOBYQA incorrect indexing
Date Sat, 12 Jul 2014 15:57:05 GMT

    [ https://issues.apache.org/jira/browse/MATH-1137?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14059831#comment-14059831
] 

Nigel Goodwin edited comment on MATH-1137 at 7/12/14 3:55 PM:
--------------------------------------------------------------

Nothing ever crashed, and I don't even know whether that bit of code is ever exercised, but
I will take up the challenge and see if I can make some tests with difficult functions (e.g.
Rosenbrook).

I met Prof Powell many decades ago (I am aged 60, which is probably 3x the age of the average
contributor to Apache) and he never claimed to be a good programmer.

http://iridia.ulb.ac.be/IridiaTrSeries/link/IridiaTr2010-010.pdf

I have looked at the NLopt C version and have little confidence in it....particularly the
rescue code, which seems to not be there in the apache java version (probably a good thing).

One other thing I noticed in Apache is line 1710 and 1744 where there is
{code}
  final int ih = nfx * (nfx + 1) / 2 - 1;
{code}

so 1 is subtracted.

All I am doing in my suggested correction of
{code}
                       double curv = modelSecondDerivativesValues.getEntry ((j + 1 + (j +1)
* (j+ 1)) / 2 - 1);
{code}

is to do the same subtraction of 1 and adjust because j starts from 0.

How do I post to the "dev" ML?




was (Author: essence):
Nothing ever crashed, and I don't even know whether that bit of code is ever exercised, but
I will take up the challenge and see if I can make some tests with difficult functions (e.g.
Rosenbrook).

I met Prof Powell many decades ago (I am aged 60, which is probably 3x the age of the average
contributor to Apache) and he never claimed to be a good programmer.

http://iridia.ulb.ac.be/IridiaTrSeries/link/IridiaTr2010-010.pdf

I have looked at the NLopt C version and have little confidence in it....particularly the
rescue code, which seems to not be there in the apache java version (probably a good thing).

One other thing I noticed in Apache is line 1710 and 1744 where there is
{code}
  final int ih = nfx * (nfx + 1) / 2 - 1;
{code}

so 1 is subtracted.

All I am doing in my suggested correction of
{curve}
                       double curv = modelSecondDerivativesValues.getEntry ((j + 1 + (j +1)
* (j+ 1)) / 2 - 1);
{curve}

is to do the same subtraction of 1 and adjust because j starts from 0.

How do I post to the "dev" ML?



> BOBYQA incorrect indexing
> -------------------------
>
>                 Key: MATH-1137
>                 URL: https://issues.apache.org/jira/browse/MATH-1137
>             Project: Commons Math
>          Issue Type: Bug
>    Affects Versions: 3.3
>            Reporter: Nigel Goodwin
>




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

Mime
View raw message