A Multi-Resolution Functional ANOVA Model for Large-Scale and Many-Dimensional Non-Linear Regression

Simulation has become an increasingly widespread technique for studying real systems in science and engineering for which physical experimentation is difficult or expensive. As computational capabilities have continued to grow simulations have become both large-scale, in terms of number of runs, and high-dimensional, in terms of input size. Here, the speaker proposes a multi-resolution functional ANOVA model for emulating large-scale and high-dimensional simulations. The model is particularly well-suited to high-dimensional data with lower-dimensional structure. Techniques for tuning parameter selection and confidence interval construction are proposed.