Topological Obstructions and How to Avoid Them
Babak Esmaeili, Robin Walters, Heiko Zimmermann, Jan-Willem van de Meent
Abstract
Incorporating geometric inductive biases into models can aid interpretability and generalization, but encoding to a specific geometric structure can be challenging due to the imposed topological constraints. In this paper, we theoretically and empirically characterize obstructions to training encoders with geometric latent spaces. We show that local optima can arise due to singularities (e.g. self-intersection) or due to an incorrect degree or winding number. We then discuss how normalizing flows can potentially circumvent these obstructions by defining multimodal variational distributions. Inspired by this observation, we propose a new flow-based model that maps data points to multimodal distributions over geometric spaces and empirically evaluate our model on 2 domains. We observe improved stability during training and a higher chance of converging to a homeomorphic encoder.
BibTeX
@inproceedings{
esmaeili2023topological,
title={Topological Obstructions and How to Avoid Them},
author={Babak Esmaeili and Robin Walters and Heiko Zimmermann and Jan-Willem van de Meent},
booktitle={Thirty-seventh Conference on Neural Information Processing Systems},
year={2023},
url={https://openreview.net/forum?id=1tviRBNxI9}
}