Riemannian Neural SDE: Learning Stochastic Representations on Manifolds
Abstract
In recent years, the neural stochastic differential equation (NSDE) has gained attention for modeling stochastic representations with great success in various types of applications. However, it typically loses expressivity when the data representation is manifold-valued. To address this issue, we suggest a principled method for expressing the stochastic representation with the Riemannian neural SDE (RNSDE), which extends the conventional Euclidean NSDE. Empirical results for various tasks demonstrate that the proposed method significantly outperforms baseline methods.
Cite
Text
Park et al. "Riemannian Neural SDE: Learning Stochastic Representations on Manifolds." Neural Information Processing Systems, 2022.Markdown
[Park et al. "Riemannian Neural SDE: Learning Stochastic Representations on Manifolds." Neural Information Processing Systems, 2022.](https://mlanthology.org/neurips/2022/park2022neurips-riemannian/)BibTeX
@inproceedings{park2022neurips-riemannian,
title = {{Riemannian Neural SDE: Learning Stochastic Representations on Manifolds}},
author = {Park, Sung Woo and Kim, Hyomin and Lee, Kyungjae and Kwon, Junseok},
booktitle = {Neural Information Processing Systems},
year = {2022},
url = {https://mlanthology.org/neurips/2022/park2022neurips-riemannian/}
}