Kolmogorov-Arnold Networks (KANs) have emerged as a groundbreaking development in deep learning, offering a more interpretable and accurate alternative to traditional Multi-Layer Perceptrons (MLPs). Developed by researchers from MIT, Caltech, Northeastern, and the NSF Institute for AI and Fundamental Interactions, KANs utilize a new architecture that replaces linear weight matrices with learnable 1D functions parameterized as splines. This design allows for the summing of incoming signals without applying non-linearities, potentially enhancing the expressiveness and efficiency of neural networks. The networks are based on the Kolmogorov-Arnold representation theorem, contrasting with MLPs that rely on the universal approximation theorem. Ziming Liu and his team have contributed significantly to the development and understanding of KANs.