ProtoGIB-Workload: Learning Workload-Specific Neural Topology Prototypes across Subjects

2026-08-11

AI Summary

This study addresses the poor generalization of models in cross-subject EEG-based mental workload recognition, which stems from inter-subject variability and noise. To this end, the authors propose ProtoGIB-Workload, a novel framework that integrates Stochastic Graph Information Bottleneck (SGIB) to compress task-relevant subgraphs and introduces a Class-conditional Topological Stabilizer (CTS). The CTS leverages fixed electrode coordinates to constrain graph generation probabilities, thereby eliminating subject-specific spurious edges and extracting consistent neural topological prototypes across subjects. Evaluated via leave-one-subject-out cross-validation on three EEG datasets, the method significantly outperforms existing approaches, achieving an average Macro-F1 improvement of 5.15% (up to 6.34%) and successfully uncovering stable cross-subject functional connectivity patterns.

Abstract

Reliable electroencephalography (EEG)-based mental workload recognition is crucial for adaptive human-centered systems, yet practical deployment requires models to generalize to users unseen during training. Although functional connectivity graphs are widely adopted to capture workload-related neural interactions, they inherently entangle task-relevant structures with subject-specific physiological traits and sample-level noise. This entanglement often leads models to learn structural shortcuts, severely degrading cross-subject generalization. To address this, we propose ProtoGIB-Workload, a novel framework that explicitly regularizes and aligns graph structures for subject-independent workload recognition. Our approach introduces a Stochastic Graph Information Bottleneck (SGIB) to compress dense correlation priors into compact, task-relevant subgraphs, filtering out input-related redundancy. Crucially, to prevent the retention of subject-specific spurious edges, we propose a Class-Conditional Topology Stabilizer (CTS). Leveraging the fixed electrode coordinates of EEG data, CTS operates directly on graph-generation probabilities to encourage consistent edge-generation statistics across different subjects sharing the same workload class. Extensive experiments on two public EEG workload datasets and one in-house EEG cognitive load dataset of air traffic controllers under strict leave-one-subject-out (LOSO) protocols demonstrate that ProtoGIB-Workload significantly outperforms state-of-the-art temporal and graph-based baselines, improving the cross-subject Macro-F1 score by an average of 5.15% (up to 6.34%). Further analyses confirm that our method successfully extracts stable, cross-subject consistent neural connectivity patterns.

Problem

• mental workload recognition
• cross-subject generalization
• functional connectivity graphs
• subject-specific traits
• neural topology

Innovation

• Stochastic Graph Information Bottleneck
• Class-Conditional Topology Stabilizer
• cross-subject generalization
• functional connectivity graph
• EEG workload recognition