SPICED: A Synaptic Homeostasis-Inspired Framework for Unsupervised Continual EEG Decoding

Yangxuan Zhou, Sha Zhao, Jiquan Wang, Haiteng Jiang, Shijian Li,, Tao Li, Gang Pan · 2025

Read SPICED: A Synaptic Homeostasis-Inspired Framework for Unsupervised Continual EEG Decoding by Yangxuan Zhou, Sha Zhao, Jiquan Wang, Haiteng Jiang, Shijian Li,, Tao Li, Gang Pan (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-17439.txt.

Description

Yangxuan Zhou and six co-authors borrow a brain trick to keep EEG decoders learning without labels. Their 2025 arXiv paper tackles continual decoding of brain signals, where new people arrive over time and each brain differs. A fixed decoder decays, but retraining with labels for every newcomer is impractical.

Their framework, SPICED, copies synaptic homeostasis, the process that keeps the brain both stable and plastic. A synaptic network grows as adaptation proceeds, driven by three bio-inspired mechanisms: reactivating key memories, consolidating synapses, and renormalizing them. The interplay strengthens traces tied to the task and weakens harmful memories. Helpful traces get replayed when a newcomer resembles past users, which aids adaptation. Harmful replays are suppressed, which curbs catastrophic forgetting. Tests on three EEG data sets back the design.

Catalog record

CREATORYangxuan Zhou, Sha Zhao, Jiquan Wang, Haiteng Jiang, Shijian Li,, Tao Li, Gang Pan
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE38 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-17439

Browse the full library · About the Tower of Babel