[AAAI-2026] $\textbf{S}^\textbf{3}$: Spiking Neurons as an Isolating Segmenter for Brain Signal Decoding

🚀 Introduction

This repository accompanies the AAAI-2026 paper.

Here, we leverage spiking neurons as an isolating segmenter for brain signal decoding. This segmenter considers subject- and task-level variability, thus providing adaptive segmentation for different brain signals. Most importantly, it exploits the unique reset mechanism of spiking neurons to enforce temporal pattern isolation for the generation of each segmentation point.

📦 Requirement

Key packages:

• python 3.10
• torch (any >=2.0 version, match your cuda)
• spikingjelly==0.0.0.0.14

You may install the required packages by:

pip install -r requirements.txt

🔑 Usage

1. Download data:

All datasets used are public on the internet for easy download.

2. Prepare the brain decoder:

Base weights of CBraMod should be loaded for further finetuning.

3. Process the data:

Run scripts in data_process/.

4. Run the main script:

python main.py  # train the SNN and the brain decoder iteratively
python main.py --frozen_ann  # froze the brain decoder, train the SNN only
python main.py --frozen_snn  # froze the SNN, train the brain decoder only
python main.py --eval  # froze both the SNN and the brain decoder, for evaluation

📄 Citation

If you find this code helpful, we would appreciate it if you cite our paper:

@inproceedings{zheng2026s3,
    title = {S³: {Spiking} {Neurons} as an {Isolating} {Segmenter} for {Brain} {Signal} {Decoding}},
    author = {Zheng, Qian and Chen, Ming and Zhao, Sha and Gu, Shi and Lin, Peng and Ma, De and Tang, Huajin and Pan, Gang},
    booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
    year={2026},
    url={https://ojs.aaai.org/index.php/AAAI/article/view/38869}
}