Clustering-based collocation for uncertainty propagation with multivariate correlated inputs

In this article, we propose the use of partitioning and clustering methods as an alternative to Gaussian quadrature for stochastic collocation (SC). The key idea is to use cluster centers as the nodes for collocation. In this way, we can extend the use of collocation methods to uncertainty propagation with multivariate, correlated input. The approach is particularly useful in situations where the probability distribution of the input is unknown, and only a sample from the input distribution is available. We examine several clustering methods and assess their suitability for stochastic collocation numerically using the Genz test functions as benchmark. The proposed methods work well, most notably for the challenging case of nonlinearly correlated inputs in higher dimensions. Tests with input dimension up to 16 are included. Furthermore, the clustering-based collocation methods are compared to regular SC with tensor grids of Gaussian quadrature nodes. For 2-dimensional uncorrelated inputs, regular SC performs better, as should be expected, however the clustering-based methods also give only small relative errors. For correlated 2-dimensional inputs, clustering-based collocation outperforms a simple adapted version of regular SC, where the weights are adjusted to account for input correlation.