How to set the optimization target and what is a result pool

[Mr. Neukam Johannes 0]

Hello,

 

I defined a result value and can use the parameter for evaluation. But where can I set, whether dakota shall optimize for lower or higher values?

 

After running a batch of optimizations with Dakota-Soga I can see, that I probably have not enough evaluations. Therefore I would like to keep the optimization going, but I can only use the run as a result pool. I could not find that term in the tutorials. What is it?

 

kind regards,

 

Johannes

[Stefan Wunderlich 6]

Hi Johannes,

 

I am afraid, there is currently no inbuild option to optimize for max value.

I would simpla invert the value of the evaluation (probably in an additional parameter) if I wanted to head for max value (value * -1).

We should really consider to do this internally and give the user a "clickable" option.

 

You are right, somehow the documentation for the result pool got lost on the road (not good).

Using a previous run as result pool means that a design created within the current run that has the same design variable values like a design in the result pool will not be evaluated but only linked to the result pool design in order to reduce computation time.

 

So it is a recommended option.

 

I hope that helps.

 

Cheers,

Stefan

[Mr. Neukam Johannes 0]

Thank you for the quick reply :-)

[Mr. Carlo Pasquinucci 1]

Hi Johannes,

 

CAESES try always to find the minimum.

 

The suggest workaround from Stefan in correct. Theoretically is possible to change the input file for Dakota to force to find the max and not the min. I really suggest you to use the Stefan method if you are not expert.

 

Otherwise, you can  write  sense max , in the response field in the Dakota template.

 

Cheers,

Carlo