Vendi Scoring for Materials Discovery

This talk will present the Vendi scores, a family of diversity measures for science and machine learning that are rooted in ecology and quantum mechanics. I'll first walk through the story behind the development of the Vendi scores and then present several techniques leveraging the Vendi scores for science. More specifically, I'll describe Vendi Sampling, an enhanced sampling technique that accelerates molecular simulations. I'll also describe Vendi Bayesian Optimization, a novel algorithm that leads to the discovery of thermally stable, energy-efficient MOFs with high storage capacity for NH3. Finally, I'll discuss opportunities for leveraging the Vendi scores to accelerate materials discovery.