We propose SPICED, a novel framework that abstracts and simplifies complex biological neural mechanisms into a computationally feasible continual learning approach, and apply it to continuous brain decoding tasks. (Arxiv)
SPICED comprises a novel synaptic network that enables dynamic expansion during continual adaptation through three bio-inspired neural mechanisms: (1) critical memory reactivation, which mimics brain functional specificity, selectively activates task-relevant memories to facilitate adaptation; (2) synaptic consolidation, which strengthens these reactivated critical memory traces and enhances their replay prioritizations for further adaptations and (3) synaptic renormalization, which are periodically triggered to weaken global memory traces to preserve learning capacities. The overview of SPICED as follows:
Step1: Pretrain the source model. PS: Any EEG decoding model can be chosen, e.g., EEGNet, EEGConformer.
parser.add_argument('--is_pretrain', type=bool, default=True, help='pretraining')Step2: To facilitate computing, we compute the initial features for all individuals before continual learning(utils/calculate_propotype.py). Regarding the selection of initial features, different feature extraction schemes can be designed according to specific tasks. We only provide a preliminary extraction approach as described in the paper.
Step3: Initialize the synaptic network and begin continual decoding.
parser.add_argument('--is_pretrain', type=bool, default=False, help='pretraining')If you're using this repository in your research or applications, please cite using the following BibTeX:
@article{zhou2025spiced,
title={SPICED: A Synaptic Homeostasis-Inspired Framework for Unsupervised Continual EEG Decoding},
author={Zhou, Yangxuan and Zhao, Sha and Wang, Jiquan and Jiang, Haiteng and Li, Shijian and Li, Tao and Pan, Gang},
journal={arXiv preprint arXiv:2509.17439},
year={2025}
}