Physics-informed NNs, Generative Physics and Astronomy
Active
Aim: Generative AI with built-in physical consistency
Tasks:
Develop and train models with new modalities for physical fields (e.g. symbolic and hybrid representations)
Improve SOTA architectures (Neural Operators, NeuralODE, PINNs, MPP)
Design algorithms for advanced data assimilation in PINNs, training stability (using RL), and better domain knowledge integration