Tag
This paper proposes using graph hypernetworks to explicitly represent relationships in PDEs for amortizing physics-informed neural networks, showing improved accuracy in solving coupled systems.