PRiSE-EEG: A Prior-Guided Foundation Model with Depth-Stratified Experts for Cross-Paradigm EEG Representation Learning

TL;DR

This paper introduces PRiSE-EEG, a novel EEG foundation model with depth-stratified experts guided by CKA analysis, enhancing cross-paradigm EEG representation learning and outperforming existing methods.

Contribution

The work proposes a prior-guided, CKA-calibrated MoE Transformer architecture with depth-stratified experts for improved cross-paradigm EEG modeling.

Findings

PRiSE-EEG achieves strong cross-paradigm performance on 12 benchmarks.

Expert allocation based on CKA improves over dense and uniform models.

Depth-wise analysis reveals early layers preserve cross-paradigm similarity.

Abstract

EEG foundation models aim to learn reusable representations across heterogeneous paradigms, yet existing approaches often use uniform adaptation mechanisms and are typically reported under separate downstream fine-tuning protocols. In this work, we first analyze dense EEG Transformers from two complementary perspectives. Gradient similarity across six downstream datasets reveals substantial optimization conflicts among EEG paradigms, while CKA analysis on mixed-paradigm batches shows a consistent depth-wise transition: shallow layers preserve stronger cross-paradigm similarity, whereas deeper layers become increasingly specialized. Motivated by these findings, we propose \textbf{PRiSE-EEG}, a prior-guided EEG foundation model with CKA-calibrated Depth-Stratified Experts. PRiSE-EEG forms continuous multi-channel EEG patches using weak static cortical and network priors and dynamic short-time channel interactions, then allocates shared and specialized experts across MoE Transformer blocks according to a sigmoid mapping from layer-wise CKA sharedness. This design preserves common EEG regularities in early blocks while assigning more specialized capacity to later task-specific transformations. Experiments on 12 public EEG benchmarks show strong cross-paradigm performance under matched protocols. Compact ablations further show that CKA-derived expert allocation improves over dense Transformers, uniform MoE, and manually fixed shared-specific expert ratios.