Re: question about donlp2 solver

From: Paul Schneider (paulibaer_at_uboot.com)
Date: 02/21/05 

Date: Mon, 21 Feb 2005 17:21:04 +0100

Peter Spellucci wrote:

[snip]
>
> from my experience
> stepsizeselection: no acceptable stepsize in [sigsm,sigla]
> means that there is an error in the gradient, in many cases due to
> nonsmooth behaviour of the function.
> this is especially true if up to that point anything "went good".
> sometimes the solution has already been identified (e.g. in a vertex
> this can occur "abruptly")
> but due to some noise in the evaluation the termination criteria are
> not satisfied
>

in function solchk it is possible to access gradf. Is this the last computed gradient, or is it the gradient evaluated
at the best parameter set ?

Best,

paul

Relevant Pages

  • Re: question about donlp2 solver
    ... >>donlp2 author Peter Spellucci directly. ... >>a numerical gradient supplied to egradf. ... >>In the manual it says that this termination message is usually due to too stringent termination criteria. ...
    (sci.math.num-analysis)
  • Re: question about donlp2 solver
    ... >Peter Spellucci wrote: ... >> but due to some noise in the evaluation the termination criteria are ... >in function solchk it is possible to access gradf. ... Is this the last computed gradient, or is it the gradient evaluated ...
    (sci.math.num-analysis)
Quantcast