Jump to content
Jörg

Surrogate-Based Optimization with Kriging, Neural Networks etc.

Recommended Posts

 Here are two posts that are related to surrogate-based optimization in CAESES:

 

 

The first topic is about re-using an existing data base to quickly set up and use a response surface model. In the example, several different models (kriging, neural network, polynomial functions etc.) are created and the one with the best fit (checked through cross-validation) is taken for the evaluation.

 

The second blog post is about automated shape optimization with CAESES. In the first stage, you create a small set of samples (e.g. through a Latin Hypercube Sampling) which is used for generating the response surface. Based on this model, a global optimization (genetic algorithm etc) can be run efficiently. The integrated charts also visualizes things such as the Pareto frontier.

Share this post


Link to post
Share on other sites

Joerg.

 

This is an interesting blog post that has some application for me.  Is there a tutorial project for this? Where can I get the Response Surface feature?

 

Britt

Share this post


Link to post
Share on other sites

Hi CAESES Team,

 

Please ill like to confirm something. For the Surrogate model in Dakota for SBGO, I see 2 types # Gaussian_process surf pack and Neural network.

Please,

 

1. Are these two used to generate the response and the one that best fits is taken for evaluation  or ?

2. There is a single default model used in CAESES SBGO ?

 

Thanks a lot.

 

Cheers.

Share this post


Link to post
Share on other sites

Hi Richard,

 

The default setting is the Gaussian process. The Neural Network is an alternative that you can also use: post-8-0-51796400-1522755362_thumb.png

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×
×
  • Create New...