Geometry-aware generative models for scientific data generation

Biological and biochemical data are increasingly high-dimensional and noisy, but they are often constrained to low-dimensional manifolds that encode meaningful state spaces, dynamics, and regulatory structure. This talk presents a unifying view of generative modeling for manifold-structured scientific data, focusing on methods that explicitly learn and exploit geometry rather than density alone.

The talk begins with SUGAR, a graph-based diffusion framework for geometry-aware data generation, then discusses PHATE, Neural FIM, GAGA, and RiTINI for manifold learning, Riemannian metrics, complex biological generative modes, and regulatory interaction network inference.