OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Convolutional Kolmogorov-Arnold Network (CKAN)

Convolutional Kolmogorov-Arnold Network (CKAN) is an MIT-licensed research implementation of convolutional neural-network layers that apply learnable nonlinear functions to individual input pixels rather than the standard linear convolution transformation. Its public repository includes a PyTorch example, reported image-classification comparisons, and a link to the accompanying academic paper.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Convolutional Kolmogorov-Arnold Network (CKAN) — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use CKAN when examining how learnable nonlinear transformations in convolutional layers might be incorporated into image-classification research or compared with conventional convolutions.

Who it is for

  • Computer-vision researchers exploring Kolmogorov-Arnold Network architectures.
  • Developers prototyping image classifiers with learnable nonlinear convolutional layers.
  • Readers comparing reported KAN-convolution experiments with conventional CNN designs.

Documented uses

  • Add a KAN convolutional layer to a Python model using the documented import.
  • Construct an example image-classification architecture with KAN convolutions, pooling and a linear output layer.
  • Review reported comparisons among CNN, KAN-convolution and mixed architectures.
Before you choose

Requirements & limits

The repository is public and identifies an MIT license.

  • The layer replaces the standard linear convolution transformation with learnable nonlinear functions applied to input elements.
  • The README reports more parameters per KAN convolution kernel than for a conventional convolution, so model-size comparisons matter.
  • The repository documents a PyTorch example and pip-based requirements installation.
  • Reported experiment results are tied to the configurations described in the repository; they do not guarantee results on other datasets or training setups.
  • The linked academic paper provides the mathematical explanation and experiment analysis.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up the repository and import the convolutional layer

Install the listed requirements and prepare the documented layer import in a Python project.

Before you start

  • A Git client and access to the public repository.
  • Python environment in which the repository's requirements can be installed.

Steps

  1. Clone the repository using the documented Git command and change into its directory.

  2. Install the project requirements with pip using the repository's requirements.txt file.

  3. Copy the kan_convolutional file into your project.

  4. Import KAN_Convolutional_Layer from kan_convolutional.KANConv.

Expected outcome

The project requirements are installed and the documented convolutional KAN layer is available to import in a Python project.

Guide limitations

  • The documented setup and import steps do not establish that a particular environment or downstream model will run successfully.
  • The README's example constructs an MNIST model, but this guide does not cover downloading data, training, or evaluating it.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
15 October 2026
Original website title
AntonioTepsich/Convolutional-KANs · Official homepage title · retained in its source language
How we review sourcesSuggest a correction