Prototype-Guided Non-Exemplar Continual Learning for Cross-subject EEG Decoding

TL;DR

This work tackles non-exemplar continual learning for cross-subject MI-EEG decoding under privacy constraints. It introduces ProNECL, which builds class-level prototypes from embeddings and maintains a global prototype memory updated by exponential moving average to enable knowledge transfer without raw data replay. The model combines supervised learning with prototype consistency and cross-subject alignment losses, and employs knowledge distillation across successive models to stabilize representations. Evaluations on the BCIC 2a and 2b datasets show that ProNECL achieves high final accuracy with minimal forgetting, outperforming baselines and demonstrating practical potential for privacy-preserving, cross-subject BCI continual learning.

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.

Figures (2)

• Figure 1: Overview of ProNECL for continual EEG decoding. (1) Base Phase: A feature extractor $\mathcal{F}_0$ is first pre-trained on the initial dataset $\mathcal{D}_0$ to learn domain-invariant EEG representations. Class-level prototypes are then computed from $\mathcal{D}_0$ and stored as reference anchors in the prototype memory. (2) Incremental Phase: When a new subject $\mathcal{D}_N$ arrives, the previous model $\mathcal{F}_{N-1}$ serves as the teacher to guide the training of the current model $\mathcal{F}_N$ through knowledge distillation, ensuring consistency of latent representations. The previously learned prototypes are projected into the latent space of $\mathcal{F}_N$ and used to align the new subject’s feature distribution with the established prototype space. This prototype-guided alignment, combined with distillation-based regularization, enables cross-subject adaptation and knowledge retention without exemplar replay.
• Figure 2: t-SNE of subject-invariant features on the 2a dataset (S1–S9): comparison without/with prototype guidance. Digits (1–9) indicate subject IDs, while markers “X”, “☆”, and “P” denote local and global prototypes, respectively.