Article Source

• Kolmogorov-Arnold Networks VS Regular Deep Learning

Kolmogorov-Arnold Networks VS Regular Deep Learning

• Paper: https://arxiv.org/abs/2404.19756
• Code: https://github.com/KindXiaoming/pykan

Abstract

MASSIVE idea proposed in this paper.

Kolmogorov-Arnold Networks (KANs) as promising alternatives to Multi-Layer Perceptrons (MLPs) for approximating nonlinear functions 🤯.

📌 Moving activation functions from nodes (neurons) to edges (weights)!

📌 MLPs place activation functions on neurons, but can we instead place (learnable) activation functions on weights? Yes, we KAN!