This repository contains the official implementation of "Real-World Performance of Kolmogorov-Arnold Networks: A Benchmarking Study" (under review). You can use this codebase to replicate our experiments about benchmarking KAN networks on some of the most used real-world tabular datasets.

Kolmogorov-Arnold Networks (KAN) has recently been introduced and gained much attention. In this work, we propose a benchmarking of KAN over some of the most used real-world datasets from UCI Machine Learning repository

1. Clone this repository: git clone https://github.com/eleonorapoeta/benchmarking-KAN.git
2. Install the required dependencies: pip install -r requirements.txt
3. Run the main script: python main.py

To reproduce the experiments conducted in our study, follow these steps:

After following the Quick Start guide, you can run python main.py specifying the following arguments:

• --model_name = (kan, mlp, all) depending on the model you want to test.
• --dataset_name = one of the tested datasets from UCI or yours.
• --num_epochs = epochs of training.

This project is licensed under the terms of the MIT license. See the LICENSE file for details.

Contributions, issues, and feature requests are welcome! See our Contributing Guide for more details.