Project Grant 2608431

The National Science Foundation Division of Mathematical Sciences awarded Florida State University $300,000 on August 1, 2026, under the Mathematical and Physical Sciences program (CFDA 47.049) to develop mathematical and computational methods based on generative artificial intelligence for reducing uncertainty in high-dimensional complex systems.

The project develops a training-free score filtering framework that integrates diffusion model-based generative AI with Bayesian data assimilation. Unlike standard diffusion models requiring computationally expensive offline neural network training, this framework constructs analytic score models derived from linear diffusion processes and approximates them using Gaussian mixture model structures. The research targets applications including atmospheric forecasting and energy system operations, where mathematical models must integrate incomplete and noisy observational data.

Florida State University's Sponsored Research Administration Division serves as the recipient. Performance occurs in Tallahassee, Florida, with a period of performance from August 1, 2026, through July 31, 2029. The project supports workforce development through graduate and undergraduate training in computational mathematics, uncertainty quantification, and scientific machine learning, with development of educational materials and open-source software tools. The award is classified as a Project Grant, an assistance type.