Knowledge Graph Integration for Biological Foundation Models Presentation of four distinct approaches to integrate knowledge graphs for biological foundation models. Diffusion Schrödinger Bridge Theoretical study of the Schrödinger Bridge problem & PyTorch implementation of the Diffusion Schrödinger Bridge algorithm to study convergence properties in the Gaussian case. Equivariant Diffusion for Molecule Generation in 3D Demonstration of the benefits of incorporating E(3)-equivariance in Graph Neural Networks through toy model experiments on the QM9 drugs dataset. Score-Based Generative Modeling Theoretical study of Score-Based Generative Modeling & PyTorch implementation to compare Langevin, SDE and ODE sampling methods. Also explored controlled generation techniques, including conditional generation and inpainting.
