DACE: A MATLAB Kriging Toolbox Version 2.0 - Implementation and Applications

DACE (Design and Analysis of Computer Experiments) is a specialized MATLAB toolbox that implements kriging approximation techniques for computer models. Version 2.0, released on August 1, 2002 by Søren N. Lophaven, Hans Bruun Nielsen, and Jacob Søndergaard, enables users to construct accurate surrogate models through Gaussian process regression. The toolbox implements key algorithms including maximum likelihood estimation for hyperparameter optimization and variogram analysis for spatial correlation modeling. A primary application involves building kriging approximation models using computer experiment data, creating efficient surrogates for computationally expensive simulations. The implementation includes data preprocessing functions, multiple covariance kernel options (Gaussian, Exponential, Matern), and cross-validation tools for model accuracy assessment. The toolbox's modular design allows customization of regression and correlation functions through MATLAB object-oriented programming. DACE has become an essential tool in computer modeling, featuring in engineering design optimization, financial risk analysis, and environmental simulation. Its implementation supports both ordinary and universal kriging variations, providing flexibility for different application requirements. The package includes visualization tools for model response surfaces and prediction variance analysis, facilitating comprehensive model evaluation and validation processes.