Neural Implicit Manifold Learning for Topology-Aware Generative Modelling

Brendan Leigh Ross, Gabriel Loaiza-Ganem, Anthony L. Caterini, Jesse C Cresswell

Keywords: Manifold Learning, Unsupervised Learning, Density Estimation, Topology, Differential Geometry, Generative Modelling

TL;DR: We propose a new model for probability distributions on topologically complex data manifolds which learns manifolds implicitly as the set of zeros of a neural network and then learns the distribution within using a constrained energy-based model.

Abstract: Natural data observed in $\mathbb{R}^n$ is often constrained to an $m$-dimensional manifold $\mathcal{M}$, where $m < n$. Current probabilistic models represent this manifold by mapping an $m$-dimensional latent variable through a neural network $f_\theta: \mathbb{R}^m \to \mathbb{R}^n$. Such procedures, which we call pushforward models, incur a straightforward limitation: manifolds cannot in general be represented with a single parameterization, meaning that attempts to do so will incur either computational instability or the inability to learn probability densities within the manifold. To remedy this problem, we propose to model $\mathcal{M}$ as a neural implicit manifold: the set of zeros of a neural network. To learn the data distribution within $\mathcal{M}$, we introduce constrained energy-based models, which use a constrained variant of Langevin dynamics to train and sample within a learned manifold. The resulting model can be manipulated with an arithmetic of manifolds, which allows practitioners to take unions and intersections of model manifolds. In experiments on synthetic and natural data, we show that constrained EBMs can learn manifold-supported distributions with complex topologies more accurately than pushforward models.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Generative models