Jim Thornton

Papers

Total Citations

H-Index

About

Jim Thornton is a rising figure in machine learning, best known for his pioneering contributions to score-based generative modelling. His most-cited work, "Riemannian Score-Based Generative Modelling" (2022), has already garnered 26 citations, establishing him as a key innovator in extending generative models beyond flat Euclidean spaces. Thornton’s research focuses on developing diffusion-based generative frameworks that operate on curved, non-Euclidean manifolds—a critical advancement for applications in molecular geometry, robotics, and 3D shape analysis. By generalising the noising and denoising processes to Riemannian manifolds, he has opened new avenues for generating complex, structured data that respects underlying geometric constraints. His work bridges the gap between theoretical geometry and practical generative AI, offering elegant solutions to problems where traditional Euclidean assumptions fall short. Thornton’s contributions are particularly notable for their mathematical rigour and potential to transform fields requiring manifold-valued data generation. With his early career already marked by high-impact, forward-looking research, Thornton is poised to become a leading voice in geometric deep learning and generative modelling.