English

Autoencoding Dynamics: Topological Limitations and Capabilities

Machine Learning 2025-11-12 v2 Dynamical Systems

Abstract

Given a "data manifold" MRnM\subset \mathbb{R}^n and "latent space" R\mathbb{R}^\ell, an autoencoder is a pair of continuous maps consisting of an "encoder" E ⁣:RnRE\colon \mathbb{R}^n\to \mathbb{R}^\ell and "decoder" D ⁣:RRnD\colon \mathbb{R}^\ell\to \mathbb{R}^n such that the "round trip" map DED\circ E is as close as possible to the identity map \mboxidM\mbox{id}_M on MM. We present various topological limitations and capabilites inherent to the search for an autoencoder, and describe capabilities for autoencoding dynamical systems having MM as an invariant manifold.

Cite

@article{arxiv.2511.04807,
  title  = {Autoencoding Dynamics: Topological Limitations and Capabilities},
  author = {Matthew D. Kvalheim and Eduardo D. Sontag},
  journal= {arXiv preprint arXiv:2511.04807},
  year   = {2025}
}
R2 v1 2026-07-01T07:25:21.465Z