jaxKAN - GitLite

KO | EN

gitlite — search

#llm #go #android #machine-learning #rust #docker #macos #java #awesome #cli #vue #security #openai #open-source

jaxKAN

(★ 112)

Kolmogorov–Arnold Networks built on JAX

File Explorer

Download Latest Version (.zip)

- - publish-to-pypi.yml
  - test.yml
- - index.rst
  - Tutorial 1 - Getting Started.ipynb
  - Tutorial 10 - Advanced Models for PDEs.ipynb
  - Tutorial 11 - Optimizer Utilities.ipynb
  - Tutorial 2 - Layer Types.ipynb
  - Tutorial 3 - Initialization Types.ipynb
  - Tutorial 4 - Classification Example.ipynb
  - Tutorial 5 - DIY KANs.ipynb
  - Tutorial 6 - PIKANs.ipynb
  - Tutorial 7 - Adaptive State Transition.ipynb
  - Tutorial 8 - Adaptive PIKANs.ipynb
  - Tutorial 9 - Inverse PDE Problem.ipynb
- about.rst
- api-reference.rst
- citing.rst
- conf.py
- contributing.rst
- index.rst
- installation.rst
- - __init__.py
  - BaseGrid.py
  - RBFGrid.py
  - SplineGrid.py
- - __init__.py
  - Chebyshev.py
  - Dense.py
  - Fourier.py
  - Legendre.py
  - RBF.py
  - Sine.py
  - Spline.py
  - utils.py
- - __init__.py
  - ActNet.py
  - KAN.py
  - KKAN.py
  - RGAKAN.py
  - utils.py
- - __init__.py
  - adaptive.py
  - pde.py
  - sampling.py
  - utils.py
- __init__.py
- - README.md
- - paper.bib
  - paper.md
- - test_chebyshev_layer.py
  - test_dense_layer.py
  - test_fourier_layer.py
  - test_layer_utils.py
  - test_legendre_layer.py
  - test_rbf_layer.py
  - test_sine_layer.py
  - test_spline_layer.py
- - test_Act.py
  - test_KAN.py
  - test_KKAN.py
  - test_model_utils.py
  - test_RGAKAN.py
- - test_adaptive.py
  - test_pde.py
  - test_pikan_utils.py
  - test_sampling.py
- conftest.py
.gitattributes
.gitignore
.readthedocs.yaml
LICENSE
pyproject.toml
README.md

# Installation Guide

1. Get the code

git clone https://github.com/srigas/jaxKAN

Downloads the entire project code from GitHub to your computer.

cd jaxKAN

Moves into the project folder you just downloaded.

2. Python

Easy Recommended

Prerequisites

Git Needed to download the project code from GitHub.
Python 3 On Windows, be sure to check 'Add Python to PATH' during installation.

pip install .

Installs the package published on PyPI directly — no need to clone the source.

jupyter notebook

Launches Jupyter in your browser so you can open and run the notebook (.ipynb) files.

✅ If it runs without errors and prints output in the terminal, it worked.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

[![DOI](https://joss.theoj.org/papers/10.21105/joss.07830/status.svg)](https://doi.org/10.21105/joss.07830)
[![Doc](https://img.shields.io/badge/docs-dev-blue.svg)](https://jaxkan.readthedocs.io/)
[![License](https://img.shields.io/github/license/srigas/jaxkan)](https://github.com/srigas/jaxKAN/blob/main/LICENSE)
[![Run Tests](https://github.com/srigas/jaxKAN/actions/workflows/test.yml/badge.svg)](https://github.com/srigas/jaxKAN/actions/workflows/test.yml)
[![PyPI version](https://img.shields.io/pypi/v/jaxkan.svg)](https://pypi.org/project/jaxkan/)

# jaxKAN

jaxKAN is a Python package designed to enable the training of Kolmogorov-Arnold Networks (KANs) using the JAX framework. Built on Flax's NNX module, jaxKAN provides a collection of KAN layers that serve as foundational building blocks for various KAN architectures, such as the EfficientKAN and the ChebyKAN. While it includes standard features like initialization and forward pass methods, the KAN class in jaxKAN introduces an `extend_grids` method, which facilitates the extension of the grids for all layers in the network, irrespective of how those grids are defined. For instance, in the case of ChebyKAN, where a traditional grid concept doesn't exist, the method extends the order of the Chebyshev polynomials utilized in the model.

## Documentation

Extensive documentation on jaxKAN, including installation & contributing guidelines, API reference and tutorials, can be found [here](https://jaxkan.readthedocs.io/).

## Contributing

We warmly welcome community contributions to jaxKAN! For details on the types of contributions that will help jaxKAN evolve, as well as guidelines on how to contribute, visit [this](https://jaxkan.readthedocs.io/en/latest/contributing.html) page of our documentation.

## Citation

If you utilized `jaxKAN` for your own academic work, please use the following citation:

```
@article{Rigas2025,
      author = {Rigas, Spyros and Papachristou, Michalis},
      title = {jax{KAN}: A unified {JAX} framework for {K}olmogorov-{A}rnold Networks},
      journal = {Journal of Open Source Software},
      year = {2025},
      volume = {10},
      number = {108},
      pages = {7830},
      doi = {10.21105/joss.07830}
}
```

If you have used jaxKAN in your research for PIKAN-related applications or theoretical developments, please consider also citing the paper that originally introduced jaxKAN for these tasks:

```
@article{10763509,
      author = {Rigas, Spyros and Papachristou, Michalis and Papadopoulos, Theofilos and Anagnostopoulos, Fotios and Alexandridis, Georgios},
      title = {Adaptive Training of Grid-Dependent Physics-Informed {K}olmogorov-{A}rnold Networks}, 
      journal = {IEEE Access},
      year = {2024},
      volume = {12},
      pages = {176982-176998},
      doi = {10.1109/ACCESS.2024.3504962}
}
```