SAC-KAN

May 26, 2026 · View on GitHub

This is the github repo for the paper "Enhancing Off-Policy Method SAC with KAN for Continuous Reinforcement Learning".

Implementation of SAC-KAN algorithm on mujoco environment，such as Hopper-v5, Halfcheeth-v5, Swimmer-v5.

Based on the Stable Baselines3.

"You can add more models simply by adding them to the Policies class (check the network.py file)."

"All hyper-parameters are overridable. To get started, see the main method in the SacKAN.py file."

Setup

  1. Clone the repository:

    git clone https://github.com/alibayeh/KAN-SAC

  2. Navigate to the project directory:

    cd KAN-SAC

  3. Install the required Python packages:

    pip install -r requirements.txt

Citation

If you use this code in your research, please cite the following paper:

@InProceedings{bayeh2025SacKan,
    author="Bayeh, Ali and Mouhoub, Malek and Sadaoui, Samira",
    editor="Hadjali, Allel and Maiorana, Emanuele and Gusikhin, Oleg and Sansone, Carlo",
    title="Enhancing Off-Policy Method SAC with KAN for Continuous Reinforcement Learning",
    booktitle="Deep Learning Theory and Applications",
    year="2025",
    publisher="Springer Nature Switzerland",
    address="Cham",
    pages="229--239",
    isbn="978-3-032-04339-9"
}