New paper published in Computers & Structures
M. Moustapha and B. Sudret published a new paper on the approximation of non-smooth functions using a three-stage approach combining clustering, classification and regression.
The paper proposes a new approach for approximating non-stationary and discontinuous functions. The proposed methodology combines a sequence of well-known machine learning/surrogate modelling techniques: Dirichlet process mixture models for clustering, support vector machines for classification and Gaussian process modelling for regression. The validation examples show that the method is very efficient in properly identifying the localized behaviours of the function to approximate and ultimately yields more accurate results compared to traditional surrogate modelling approaches.
For more information, please follow this external pagelink for the publication and this link for the associated report on our internal archive.
-