comp.lang.idl-pvwave archive: archive

Home » Public Forums » archive » MPfit question

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

Re: MPfit question [message #62738 is a reply to message #62623]

Thu, 02 October 2008 04:39

Wox
Messages: 184
Registered: August 2006

Senior Member

On 01 Oct 2008 13:16:10 -0400, Craig Markwardt
<craigmnet@REMOVEcow.physics.wisc.edu> wrote:

>
> Wox <nomail@hotmail.com> writes:
>
>> On 30 Sep 2008 11:49:46 -0400, Craig Markwardt
>> <craigmnet@REMOVEcow.physics.wisc.edu> wrote:
>>
>>> If you look at the code, the value of ALPHA is adjusted so that, at
>>> the next iteration, a parameter will exactly touch its boundary,
>>> within a small tolerance. At that point, the parameter will be
>>> considered fixed, and will no longer enter into the calculation of the
>>> value of ALPHA. [*] Thus, the step *is* adaptive, it just doesn't
>>> happen in a single iteration.
>>
>> I'm sorry, but I don't see how it does this. ALPHA is adjusted and
>> immediatly used (see below). In the next iteration, the increments are
>> calculated again by mpfit_lmpar and used again to calculate ALPHA,
>> whether the param was at the limit in the previous iteration or not.
>
> That is not correct. Please search for 'zeroing the derivatives of
> pegged parameters'. Once a parameter is pegged at a boundary in the
> previous iteration, it no longer contributes to the congugate gradiate
> solution because its derivatives have been zeroed.

Ah yes, I see now. Thanks for your comments. I must say it's quite an
impressive and complicated piece of code. I'm having a hard time
understanding the convergence criteria and the calculation of the
LM-step. Can you recommend some reference where I can find more
details? I sure have some books explaining NLLS-refinement but they
don't seem to mention some of the criteria I see in your code. Also
the calculation of the LM-step is done differently each time I open a
new book :-).

Report message to a moderator

[Message index]

		MPfit question By: Wox on Tue, 30 September 2008 07:38
		Re: MPfit question By: Craig Markwardt on Thu, 02 October 2008 08:59
		Re: MPfit question By: Wox on Thu, 02 October 2008 04:39
		Re: MPfit question By: Craig Markwardt on Wed, 01 October 2008 10:16
		Re: MPfit question By: Wox on Wed, 01 October 2008 01:31
		Re: MPfit question By: Craig Markwardt on Tue, 30 September 2008 08:49
		Re: MPFIT question By: Craig Markwardt on Tue, 13 January 2009 22:41
		Re: MPFIT question By: j.coenia@gmail.com on Wed, 14 January 2009 07:54
		Re: MPFIT question By: j.coenia@gmail.com on Thu, 15 January 2009 08:17
		Re: MPFIT question By: Craig Markwardt on Fri, 16 January 2009 22:10
		Re: MPFIT question By: Craig Markwardt on Wed, 14 January 2009 23:52
		Re: MPFIT question By: pgrigis on Wed, 14 January 2009 14:08
		Re: MPFIT question By: j.coenia@gmail.com on Wed, 14 January 2009 13:37
		Re: MPFIT question By: Craig Markwardt on Wed, 14 January 2009 08:23

Previous Topic:	Trouble writing very large files
Next Topic:	Recording batch commands

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Fri Oct 10 04:11:29 PDT 2025

Total time taken to generate the page: 0.24216 seconds