Thread Subject: Maximum number of iterations exceeded

Subject: Maximum number of iterations exceeded

From: christos

Date: 17 Nov, 2009 17:18:03

Message: 1 of 6

Dear all

I want to minimize two cost-functions in an alternating fashion. By alternating I mean that I take some steps towards the minimization of the first function followed by some steps towards the minimization of the second. Effectively these steps are the iterations of the optimizer. I set this to a small number (like 5) however I get the following:

Maximum number of iterations exceeded;
   increase options.MaxIter.

Is this some kind of embedded Matlab functionality that if it thinks that it can minimize further, it simply ignores the MaxIter option?
If that is the case, is there any workaround in order to enforce the MaxIter?

thanks
christos

Subject: Maximum number of iterations exceeded

From: Matt

Date: 17 Nov, 2009 17:27:18

Message: 2 of 6

"christos " <spamsaver3000@googlemail.com> wrote in message <hdulsb$pvj$1@fred.mathworks.com>...
> Dear all
>
> I want to minimize two cost-functions in an alternating fashion. By alternating I mean that I take some steps towards the minimization of the first function followed by some steps towards the minimization of the second. Effectively these steps are the iterations of the optimizer. I set this to a small number (like 5) however I get the following:
>
> Maximum number of iterations exceeded;
> increase options.MaxIter.
>
> Is this some kind of embedded Matlab functionality that if it thinks that it can minimize further, it simply ignores the MaxIter option?
> If that is the case, is there any workaround in order to enforce the MaxIter?
====

This message is simply to warn you that MaxIter has been reached. Typically, this is a bad thing - most people reach MaxIter unintentionally because the function they are trying to minimize is poorly conditioned and not enough iterations were run to reach a more savvy convergence criterion.

However, your case is not typical: you fully expect and desire the iterations to terminate after a certain number. So I think you can ignore it.

Subject: Maximum number of iterations exceeded

From: christos

Date: 17 Nov, 2009 19:13:19

Message: 3 of 6

thanks for answering matt

the problem is that i can't ignore it because after the "iteration exceeded" message, matlab seems to take the initiative of increasing the iterations...

my problem is that the MaxIter is not strictly enforced

any suggestions/workarounds guys?

Subject: Maximum number of iterations exceeded

From: Matt

Date: 17 Nov, 2009 19:19:20

Message: 4 of 6

"christos " <spamsaver3000@googlemail.com> wrote in message <hduskf$4ej$1@fred.mathworks.com>...
> thanks for answering matt
>
> the problem is that i can't ignore it because after the "iteration exceeded" message, matlab seems to take the initiative of increasing the iterations...

What evidence do you see of that?

Subject: Maximum number of iterations exceeded

From: christos

Date: 17 Nov, 2009 19:53:03

Message: 5 of 6

hmmm

no evidence at all -
sorry about that.... the optimizer was running under a hidden loop (scale space registration - well not "hidden", but i ve missed it) so it seemed like it seemed like going for further iterations

lol... i think i ve missed this big time
sorry and thanks for your time :)

Subject: Maximum number of iterations exceeded

From: JUAN Tapia

Date: 31 Aug, 2010 16:08:04

Message: 6 of 6

Hi
the solution is change QP for SMO in SVM train
for example
result=svmtrain(A,B,'METHOD','SMO',KERNEL_FUNCTION','rbf',rbf_sigma',16);

this is all

reagrds

Tags for this Thread

Everyone's Tags:

Add a New Tag:

Separated by commas
Ex.: root locus, bode

What are tags?

A tag is like a keyword or category label associated with each thread. Tags make it easier for you to find threads of interest.

Anyone can tag a thread. Tags are public and visible to everyone.

Tag Activity for This Thread
Tag Applied By Date/Time
svm train error JUAN Tapia 31 Aug, 2010 12:09:18
optimization christos 17 Nov, 2009 12:19:07
rssFeed for this Thread

Contact us at files@mathworks.com