GitHunt
XA

Xatta-Trone/awesome-kan

A curated collection of KAN (Kolmogorov-Arnold Network) resources—libraries, projects, tutorials, papers, and more—for researchers and developers in the field.

Awesome KAN (Kolmogorov-Arnold Network)

Contributions GitHub stars

A carefully curated collection of exceptional libraries, projects, tutorials, research papers, and other resources focused on Kolmogorov-Arnold Networks (KAN). This repository serves as a well-structured and thorough resource hub, designed to support and guide researchers and developers exploring the field of KAN.

Our repository is automatically updated with the latest KAN-related research papers from arXiv, ensuring that users have access to the most up-to-date advancements in the field.

We are pleased to announce that our comprehensive review paper, "A Survey on Kolmogorov-Arnold Network", has been accepted to ACM Computing Surveys (ACM CS). This paper explores the theoretical foundations, architectural advances such as FastKAN, T-KAN, and PDE-KAN, optimization strategies, and practical applications across multiple domains. It can be accessed at https://doi.org/10.1145/3743128.

Whether you are a researcher, developer, or enthusiast, this collection provides a centralized hub for everything related to KAN, now featuring original peer-reviewed contributions to the field.

Last Updated

March 11, 2026 at 01:05:22 AM UTC

Table of Contents

Theorem

Papers (660)

Library

  • xKAN - An advanced implementation of KAN incorporating multiple basis functions such as B-Splines, Fourier, Chebyshev, and Wavelets.
  • RBF-KAN - A PyTorch-based Kolmogorov-Arnold Network utilizing Radial Basis Functions (RBF) for enhanced performance.
  • GraphKAN - A specialized KAN library adapted for Graph Neural Networks (GNNs), enabling graph-based learning tasks.
  • BSRBF_KAN - A hybrid approach combining B-Spline and Radial Basis Function (RBF) layers within KAN.
  • KANX - A highly efficient JAX-based implementation of KAN for accelerated computations.
  • Deep-KAN - An improved and optimized implementation of the Kolmogorov-Arnold Network in PyTorch.
  • TorchKAN - Implements KAN using Legendre and monomial basis functions, achieving high accuracy on image classification tasks.
  • fKAN - A fractional version of KAN incorporating trainable Jacobi basis functions.
  • pykan - The official reference implementation of Kolmogorov-Arnold Networks.
  • FastKAN - An optimized and high-speed computation framework for KAN, significantly improving efficiency.
  • Quantum KAN - A novel KAN implementation that integrates quantum annealing optimization techniques.
  • FlashKAN - A specialized implementation of KAN with grid-size-independent computations for enhanced efficiency.
  • FourierKAN - A PyTorch-based Fourier transformation layer tailored for KAN.
  • OrthogPolyKAN - A KAN variant that utilizes orthogonal polynomials instead of conventional B-splines.
  • FasterKAN - An accelerated version of KAN featuring RSWAF basis functions, delivering faster backward passes.
  • KAN-SGAN - A semi-supervised learning implementation integrating KAN layers into Generative Adversarial Networks (GANs).
  • efficient-kan - A highly efficient, pure PyTorch implementation of KAN for deep learning applications.
  • TaylorKAN - A KAN model utilizing Taylor series in place of Fourier transformations.
  • Wav-KAN - Implements wavelet-based transformations within the KAN framework.
  • ChebyKAN - A variation of KAN leveraging Chebyshev polynomials instead of traditional activation functions.
  • Vision-KAN - A KAN implementation designed for Vision Transformers, significantly improving performance on image classification tasks.
  • KAN-Conv2D - A drop-in replacement for Conv2D layers using KAN, built on multiple KAN implementations.
  • JacobiKAN - A KAN implementation replacing B-Splines with Jacobi polynomials to enhance approximation accuracy.
  • efficient-kan-jax - A JAX-based port of efficient-kan for optimized execution on modern hardware.
  • cuda-Wavelet-KAN - A CUDA-optimized KAN implementation specifically for wavelet-based transformations.
  • ConvKAN3D - A 3D convolutional implementation of KAN for applications requiring volumetric data processing.
  • x-KANeRF - An extension of KAN applied to Neural Radiance Fields (NeRF) for improved scene rendering.
  • KolmogorovArnold.jl - A high-performance Julia implementation of KAN featuring RBF and RSWAF basis functions.
  • KAN: Kolmogorov–Arnold Networks in MLX for Apple Silicon - A specialized KAN implementation optimized for Apple's MLX framework.
  • Torch Conv KAN - Implements convolutional KAN layers, with support for 1D, 2D, and 3D convolutions.
  • keras_efficient_kan - A TensorFlow/Keras implementation of efficient KAN supporting multi-backend compatibility.
  • kanrl - A reinforcement learning-focused implementation of KAN, designed for policy learning tasks.
  • jaxKAN - A JAX implementation of KAN with complete support for regularization techniques.
  • GraphKAN - A graph-structured KAN model designed for advanced feature extraction.
  • kan-polar - A MATLAB-based KAN implementation for researchers and engineers.
  • seydi1370/Basis_Functions - A benchmark study analyzing the performance of 18 different polynomial basis functions for KAN.
  • FCN-KAN - A KAN model featuring modified activations using a fully connected network structure.
  • TKAN - Implements Temporal Kolmogorov-Arnold Networks (TKAN) for time-series modeling.
  • FluxKAN.jl - A user-friendly Julia-based KAN implementation built using the Flux ML framework.
  • ikant - A lightweight C++ implementation of KAN with built-in visualization tools.
  • kamo - A KAN framework implemented in Mojo, optimized for high-performance computing.
  • KAN-benchmarking - A benchmarking suite evaluating the efficiency of different KAN implementations in terms of speed and memory usage.
  • CF-KAN - A KAN-based collaborative filtering algorithm designed for recommender systems.
  • TKAT - A Kolmogorov-Arnold Transformer (KAT) model optimized for time-series forecasting.
  • KAN4Graph - An application of KAN for Graph Neural Networks (GNNs) and graph-based machine learning tasks.

Tutorial

Written Tutorials

Video Tutorials

Contributing

We welcome your contributions! Please follow these steps to contribute:

  1. Fork the repo.
  2. Create a new branch (e.g., feature/new-kan-resource).
  3. Commit your changes to the new branch.
  4. Create a Pull Request, and provide a brief description of the new resources.

Please make sure that the resources you add are relevant to the field of Kolmogorov-Arnold Network. Before contributing, take a look at the existing resources to avoid duplicates.

License

This work is licensed under a Creative Commons Attribution 4.0 International License.

Credits

This curated collection of resources, tutorials, and libraries on Kolmogorov-Arnold Networks (KANs) is inspired by the excellent work done in Awesome KAN.
We acknowledge and appreciate the contributors of the original repository for compiling a vast amount of knowledge on KANs, making it more accessible to researchers, engineers, and enthusiasts.
If you find this collection useful, consider checking out the original Awesome KAN repository and contributing to the ever-growing knowledge base on Kolmogorov-Arnold Networks!

Star History

Star History Chart

Xatta-Trone/awesome-kan | GitHunt