PH-KAN: Port-Hamiltonian Kolmogorov-Arnold Network

The article introduces the Port-Hamiltonian Kolmogorov-Arnold Network (PH-KAN), a novel framework for identifying nonlinear port-Hamiltonian systems using Kolmogorov-Arnold Networks (KANs). This model enhances interpretability by parameterizing key components such as the interconnection and dissipation matrices, as well as the Hamiltonian, while ensuring adherence to port-Hamiltonian constraints. This approach is significant for practitioners as it allows for clearer insights into the learned physical relationships in complex systems, improving the usability of machine learning in physics-informed modeling.