Prev: signal problem
Next: Neural Network inputs
From: Roxanna Moores on 14 Apr 2010 16:57
I am trying to optimize parameters to a simulink model with an S-function that takes the parameters that I want to optimize. When I run the parameter optimization, the cost function does not change for the various iterations, and my model does not show any changes from the very first prediction that takes the initial parameter estimations. It is as if the new estimates for the parameters are not being updated in the workspace for the S-function.

Does anyone know why the paramter optimization is behaving this way?
From: Arkadiy Turevskiy on 19 Apr 2010 14:21
It probably has nothing to do with parameter optimization, but with how your model is configured. You need to check that when model is executed parameter values are not overwritten - do you have any callbacks in the model that could be overvwriting parameter values in the workspace, for example?

Have you tried changing parameter values in the workspace, running the model, and checking if it used the right parameter values?

Arkadiy


"Roxanna Moores" <roxanna.moores(a)volvo.com> wrote in message <hq5a70$oo0$1(a)fred.mathworks.com>...
> I am trying to optimize parameters to a simulink model with an S-function that takes the parameters that I want to optimize. When I run the parameter optimization, the cost function does not change for the various iterations, and my model does not show any changes from the very first prediction that takes the initial parameter estimations. It is as if the new estimates for the parameters are not being updated in the workspace for the S-function.
>
> Does anyone know why the paramter optimization is behaving this way?
From: Roxanna Moores on 19 Apr 2010 15:21
Hello,
I did try to change the parameters in the workspace, and it does change the estimation. I also did check for callbacks, I didn't find any.

I believe that either you are right and the parameters are being overwritten or the new estimtions are never being utilized somehow. I do know that in the workspace, the paramter values do not change from the initialized values.

But I'm still stumped as to how to find the source of overwriting or how to get the new estimations to be used by parameter estimation. I feel like the S-function is interfering somehow.

Roxanna

"Arkadiy Turevskiy" <arkadiy.turevskiy(a)mathworks.com> wrote in message <hqi6uh$l6f$1(a)fred.mathworks.com>...
> It probably has nothing to do with parameter optimization, but with how your model is configured. You need to check that when model is executed parameter values are not overwritten - do you have any callbacks in the model that could be overvwriting parameter values in the workspace, for example?
>
> Have you tried changing parameter values in the workspace, running the model, and checking if it used the right parameter values?
>
> Arkadiy
>
>
> "Roxanna Moores" <roxanna.moores(a)volvo.com> wrote in message <hq5a70$oo0$1(a)fred.mathworks.com>...
> > I am trying to optimize parameters to a simulink model with an S-function that takes the parameters that I want to optimize. When I run the parameter optimization, the cost function does not change for the various iterations, and my model does not show any changes from the very first prediction that takes the initial parameter estimations. It is as if the new estimates for the parameters are not being updated in the workspace for the S-function.
> >
> > Does anyone know why the paramter optimization is behaving this way?
 | 
Pages: 1
Prev: signal problem
Next: Neural Network inputs