4178249 - Accelerating design and optimization in molecular discovery: Generative models and beyond

Exploring vast search spaces with costly experimental or simulation validations is a common challenge in many real-world chemistry problems. Recent advances in generative models offer great promise for learning good priors in design and optimization problems. However, molecular discovery problems often come with varying data availabilities. In this talk, I will discuss how generative models and more general optimal transport and stochastic optimal control methods can efficiently and effectively leverage data to accelerate molecular design, optimization, transition state search and transition path sampling.