Additive models in high dimensions
Data Structures and Algorithms
2009-11-17 v2
Abstract
We discuss some aspects of approximating functions on high-dimensional data sets with additive functions or ANOVA decompositions, that is, sums of functions depending on fewer variables each. It is seen that under appropriate smoothness conditions, the errors of the ANOVA decompositions are of order for approximations using sums of functions of up to variables under some mild restrictions on the (possibly dependent) predictor variables. Several simulated examples illustrate this behaviour.
Cite
@article{arxiv.cs/9912020,
title = {Additive models in high dimensions},
author = {Markus Hegland and Vladimir Pestov},
journal= {arXiv preprint arXiv:cs/9912020},
year = {2009}
}
Comments
LaTeX 2e document, 21 pages, 5 figures