https://github.com/Sid2690/Deep-KAN

Better implementation of Kolmogorov Arnold Network
https://github.com/Sid2690/Deep-KAN

Last synced: about 1 month ago
JSON representation

Better implementation of Kolmogorov Arnold Network

• Host: GitHub
• URL: https://github.com/Sid2690/Deep-KAN
• Owner: Sid2690
• License: mit
• Created: 2024-05-15T15:41:08.000Z (8 months ago)
• Default Branch: main
• Last Pushed: 2024-06-09T06:53:30.000Z (7 months ago)
• Last Synced: 2024-06-09T07:48:21.618Z (7 months ago)
• Language: Jupyter Notebook
• Size: 215 KB
• Stars: 12
• Watchers: 1
• Forks: 4
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-kan - Deep-KAN - KAN.svg) (Library / Theorem)

README

        
# Kolmogorov-Arnold-Networks-KAN

I have implemented KAN using B-splines in a more efficient manner to enhance its performance. Additionally, I have thoroughly tested its performance on various datasets, including MNIST, Air Passengers, and CIFAR-10.

I will soon upload the results of all my experiments to provide a comprehensive overview of KAN's performance across different datasets.

# Deep-KAN

Deep-KAN is a Python package for implementing Kolmogorov-Arnold-Networks (KAN). 

You can find the package [here](https://pypi.org/project/Deep-KAN/).

# MLP v/s KAN

For 20 epochs with a single hidden layer(32 neurons)

(Note: both are untuned)

![image](https://github.com/sidhu2690/KAN/assets/136654152/5e0c2b8a-eb13-4110-8b2e-01809537f0f8)

# RBF-KAN

You can find the RBF implementation of KAN [here](https://github.com/sidhu2690/RBF-KAN).