Riemannian Denoising Diffusion Probabilistic Models
This method addresses generative modeling on general manifolds for applications like molecular systems, offering a more flexible approach compared to existing methods that rely on extensive geometric information.
The authors tackled the problem of generative modeling on submanifolds of Euclidean space, such as level sets of functions, by proposing Riemannian Denoising Diffusion Probabilistic Models (RDDPMs), which only require evaluating the function and its first-order derivatives, and demonstrated its capability on datasets including SO(10) and alanine dipeptide.
We propose Riemannian Denoising Diffusion Probabilistic Models (RDDPMs) for learning distributions on submanifolds of Euclidean space that are level sets of functions, including most of the manifolds relevant to applications. Existing methods for generative modeling on manifolds rely on substantial geometric information such as geodesic curves or eigenfunctions of the Laplace-Beltrami operator and, as a result, they are limited to manifolds where such information is available. In contrast, our method, built on a projection scheme, can be applied to more general manifolds, as it only requires being able to evaluate the value and the first order derivatives of the function that defines the submanifold. We provide a theoretical analysis of our method in the continuous-time limit, which elucidates the connection between our RDDPMs and score-based generative models on manifolds. The capability of our method is demonstrated on datasets from previous studies and on new datasets sampled from two high-dimensional manifolds, i.e. $\mathrm{SO}(10)$ and the configuration space of molecular system alanine dipeptide with fixed dihedral angle.