Computational Mathematics Seminar Series sponsored by the LSU Center for Computation & Technology (CCT)

Guest lecture from Yanzhao Cao, Auburn University 

Title: A training-free diffusion model for generative learning 

Abstract: In this talk, I will first present a framework for training generative models for density estimation using stochastic differential equations (SDEs). Unlike conventional diffusion models that train neural networks to learn the score function, we introduce a score-estimation method that is training-free. This approach uses mini-batch-based Monte Carlo estimators to directly approximate the score function at any spatiotemporal location while solving the ordinary differential equation (ODE) corresponding to the reverse-time SDE. Our method provides high accuracy and significant reductions in neural network training time. Algorithm development and convergence analysis will be discussed. At the end, I will present an application of the diffusion model to fusion plasma.

March 24, 2026

3:30pm DMC 1034

Refreshments will be served at 3:00pm