Brief description
The ooDACE Toolbox (Design and Analysis of Computer Experiments) is a powerful and versatile Matlab toolbox for building Kriging surrogate models of a given data set (e.g., based on computationally expensive simulations or physical experiments). Kriging is, in particular, popular for approximating (and optimizing) deterministic computer experiments. Kriging surrogate models are compact scalable regression models which can be used efficiently for design automation, parametric studies, design space exploration, optimization, yield improvement, visualization, prototyping, and sensitivity analysis. The ooDACE Toolbox provides a flexible Object Oriented Matlab implementation, easily extendable and well-suited to test and benchmark new kriging flavors. Matlab 2015a or newer is recommended.
Key features
-
Support for several variants of Kriging: simple Kriging, ordinary Kriging, universal Kriging, blind Kriging, co-Kriging,regression Kriging, etc.
-
Efficient and flexible hyperparameter optimization
-
Can handle noisy data (by including an extra parameter in the hyperparameter optimization) = regression Rriging
-
Includes tools: cross-validation, robustness-criterion value, prediction mean and variance + derivatives (e.g., for sensitivity analysis), variograms, debug plots, etc.
-
Interface compatible with the popular Matlab DACE toolbox
-
Proper Object Oriented (OO) design