Adaptive joint distribution learning
Machine Learning
2024-09-25 v5 Machine Learning
Numerical Analysis
Numerical Analysis
Abstract
We develop a new framework for estimating joint probability distributions using tensor product reproducing kernel Hilbert spaces (RKHS). Our framework accommodates a low-dimensional, normalized and positive model of a Radon--Nikodym derivative, which we estimate from sample sizes of up to several millions, alleviating the inherent limitations of RKHS modeling. Well-defined normalized and positive conditional distributions are natural by-products to our approach. Our proposal is fast to compute and accommodates learning problems ranging from prediction to classification. Our theoretical findings are supplemented by favorable numerical results.
Cite
@article{arxiv.2110.04829,
title = {Adaptive joint distribution learning},
author = {Damir Filipovic and Michael Multerer and Paul Schneider},
journal= {arXiv preprint arXiv:2110.04829},
year = {2024}
}