Towards Physically Reliable Molecular Representation Learning - UAI 2023 Oral Session 5
Explore a cutting-edge conference talk on molecular representation learning presented at the Uncertainty in Artificial Intelligence (UAI) 2023 conference. Delve into the challenges of estimating energetic properties of molecular systems for material design and discover a novel approach that combines physics-driven parameter estimation with self-supervised learning. Learn about the proposed method's effectiveness in discovering molecular structures and its potential for extrapolating to chemical reaction pathways beyond stable states. Gain insights into new evaluation schemes that go beyond traditional energy and force accuracy measurements, and understand how this research contributes to the development of physically reliable molecular representation learning.