Physics-informed neural networks
■ Adaptive sampling method for PINNs
▲ PINNs coupled with adaptive sampling techniques for efficient training
■ Case study of lid-driven cavity flow
▲ Comparison with OpenFOAM and PINNs (Re=100)
■ Multi-fidelity PINNs
▲ Proposed multi-fidelity data-guided PINNs:
remarkable extrapolation performance w.r.t. Reynolds number can be verified