Augmented Neural ODEs
Machine Learning
2019-10-29 v3 Machine Learning
Abstract
We show that Neural Ordinary Differential Equations (ODEs) learn representations that preserve the topology of the input space and prove that this implies the existence of functions Neural ODEs cannot represent. To address these limitations, we introduce Augmented Neural ODEs which, in addition to being more expressive models, are empirically more stable, generalize better and have a lower computational cost than Neural ODEs.
Keywords
Cite
@article{arxiv.1904.01681,
title = {Augmented Neural ODEs},
author = {Emilien Dupont and Arnaud Doucet and Yee Whye Teh},
journal= {arXiv preprint arXiv:1904.01681},
year = {2019}
}
Comments
NeurIPS camera ready, additional experiments, additional datasets, discussion on relation to other models