RECTOR: Masked Region-Channel-Temporal Modeling for Affective and Cognitive Representation Learning

RECTOR (Masked Region-Channel-Temporal Modeling) is an end-to-end self-supervised framework designed for affective and cognitive representation learning from EEG/sEEG data. It employs a hierarchical, block-sparse self-attention mechanism driven by Adaptive Functional Partitioning, enabling dynamic region structures and optimizing objectives through Masked Topology and Representation Learning. RECTOR achieves state-of-the-art performance in EEG emotion recognition and sEEG task-engagement classification, demonstrating robustness to missing channels and potential for large-scale pre-training, which is crucial for enhancing clinical diagnosis capabilities in neuroscience applications.