Kolmogorov-Arnold Networks (KANs) are introduced as a promising alternative to Multi-Layer Perceptrons (MLPs). KANs are based on the Kolmogorov-Arnold representation theorem, giving them strong mathematical foundations. KANs outperform MLPs in terms of model accuracy and interpretability, with examples ranging from fitting symbolic formulas to discovering physical laws. Installation of the KAN library, pykan, can be done through either pypi or github. Computation requirements vary, with larger tasks benefitting from GPU usage. The documentation, tutorials, and citation information are available for further exploration. The authors provide their contact information for any inquiries.
https://github.com/KindXiaoming/pykan