help-gsl
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Help-gsl] Possible multiroot shortcoming


From: Brian Gough
Subject: Re: [Help-gsl] Possible multiroot shortcoming
Date: Thu, 23 Aug 2007 14:12:31 +0100
User-agent: Wanderlust/2.14.0 (Africa) Emacs/22.1 Mule/5.0 (SAKAKI)

At Wed, 22 Aug 2007 08:42:15 -0700,
Mark Martin wrote:
> 
> I've been attempting to use the multiroot facility and may have discovered an 
> unpleasant feature.  The documentation states that the solvers will return an 
> error code if the function or its derivative evaluate to Inf or NaN or if the 
> iteration is not making progress toward a root.  However, it appears that the 
> solver can also fail if it calculates some components of the approximation, 
> i.e. next guess, to be NaN when it did not detect aberrant behavior in the 
> function or its derivative.  The only warning that such a failure occurred is 
> when the solver later indicates that it is not making progress toward a 
> solution.  Unfortunately, I cannot post any reasonably simple code that 
> demonstrates this problem, which is partially why I'm not submitting a formal 
> bug report.  But I thought that you might want to know in case this isn't 
> already a known feature.

I imagine that could happen if the function has some extreme behavior,
so the computed step goes to infinity or whatever, but it shouldn't
happen if the function is reasonably well behaved.

If you do find a simple example of the problem, can you submit it as a
bug report to address@hidden Thank you.

-- 
Brian Gough

Network Theory Ltd,
Publishing the GSL Manual - http://www.network-theory.co.uk/gsl/manual/





reply via email to

[Prev in Thread] Current Thread [Next in Thread]