KANLib -- An Modular, Extensible and Fast Kolmogorov-Arnold Network Implementation

KANLib is a newly released framework designed for the development and evaluation of Kolmogorov-Arnold Networks (KANs), addressing the challenges of high computational costs and feature inconsistencies in existing implementations. It integrates concepts from previous frameworks like PyKAN, EfficientKAN, and FastKAN, offering features such as adaptive grid rescaling and fine-grained customization while ensuring compatibility with PyTorch. The framework has demonstrated competitive efficiency on the California Housing benchmark, making it a valuable tool for researchers looking to explore and implement scalable KAN architectures.