deepone

KernelOnet proposes an interpretable neural operator framework that explicitly incorporates kernel functions in place of trunk networks, offering data-driven, physics-informed, and hybrid kernel strategies for PDE solving with higher accuracy and fewer parameters than DeepONet.