Technology Category
Machine Learning: Life-Long and Continual LearningCognitive Modeling & Cognitive Systems: Neural Spike CodingComputer Vision: Learning & Optimization for CV
Prototype-Guided Non-Exemplar Continual Learning for Cross-subject EEG Decoding
🤖 AI Summary
Cross-subject EEG continual decoding suffers from catastrophic forgetting of historical knowledge due to inter-subject signal variability, while existing replay-based approaches are constrained by privacy concerns and memory limitations. To address this, we propose a prototype-guided exemplar-free continual learning framework: it constructs class-level incremental global prototype memories and jointly leverages cross-subject feature alignment and knowledge distillation to preserve and transfer prior knowledge—without storing any raw EEG samples. This is the first method to achieve efficient cross-subject knowledge reuse under strict replay-free conditions. Evaluated on BCI Competition IV datasets 2a and 2b, it significantly outperforms state-of-the-art methods in both knowledge retention and model adaptability. The framework establishes a novel paradigm for privacy-sensitive, continual brain–computer interface learning.
Technology Category
Application Category
📝 Abstract
Due to the significant variability in electroencephalogram (EEG) signals across individuals, knowledge acquired from previous subjects is often overwritten as new subjects are introduced in continual EEG decoding task. Current works mainly rely on storing the historical data of seen subjects as a replay buffer to prevent forgetting. However, privacy concerns or memory constraints make keeping such data impractical. Instead, we propose a Prototype-guided Non-Exemplar Continual Learning (ProNECL)framework that preserves prior knowledge without accessing any historical EEG samples. ProNECL constructs class-level prototypes to summarize discriminative representations from each subject and incrementally aligns new feature spaces with the global prototype memory through cross-subject feature alignment and knowledge distillation. Validated on the BCI Competition IV 2a and 2b datasets, our framework effectively balances knowledge retention and adaptability, achieving superior performance in cross-subject continual EEG decoding tasks.
Problem
Research questions and friction points this paper is trying to address.
Addresses catastrophic forgetting in cross-subject EEG decoding
Eliminates need for storing historical EEG data due to privacy constraints
Aligns new feature spaces with prototypes to preserve prior knowledge
Innovation
Methods, ideas, or system contributions that make the work stand out.
Constructs class-level prototypes for discriminative representations
Aligns new feature spaces with global prototype memory
Uses cross-subject feature alignment and knowledge distillation
🔎 Similar Papers
No similar papers found.
