AntonioTepsich/Convolutional-KANs

This project extends the idea of the innovative architecture of Kolmogorov-Arnold Networks (KAN) to the Convolutional Layers, changing the classic linear transformation of the convolution to learnable non linear activations in each pixel.

45
/ 100
Emerging

This project offers an alternative way to process images in deep learning models by replacing traditional convolutional layers with a new type of layer called a Convolutional Kolmogorov-Arnold Network (CKAN). Instead of standard linear operations, CKANs use learnable, non-linear functions at each pixel, which can help models better understand complex patterns in visual data. It's designed for machine learning researchers and practitioners who are experimenting with novel neural network architectures for image-related tasks.

917 stars. No commits in the last 6 months.

Use this if you are a machine learning researcher or practitioner interested in exploring new neural network architectures for image classification and want to leverage the theoretical benefits of Kolmogorov-Arnold Networks in convolutional settings.

Not ideal if you need models with extremely fast training times or if you prioritize minimizing parameter count for image classification, as CKANs currently have longer training times and more parameters per layer than traditional CNNs.

image-classification neural-networks deep-learning-research computer-vision model-architecture
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 10 / 25
Maturity 16 / 25
Community 19 / 25

How are scores calculated?

Stars

917

Forks

97

Language

Jupyter Notebook

License

MIT

Last pushed

Apr 08, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/AntonioTepsich/Convolutional-KANs"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.