EEG-based cognition-aware task classification and scheduling using enhanced fuzzy transition modeling.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Cognitive Science

Background:

• Cognitive state modeling (CSM) traditionally uses classification, limiting real-world adaptive applications.
• Existing methods struggle to link brain state classifications to task-level decision-making.
• There's a need for models that treat cognition as a continual process for better integration into decision-support systems.

Purpose of the Study:

• To propose a neurosymbolic model for cognitive state modeling as a continual latent process.
• To develop a framework that integrates neural representation learning with symbolic reasoning for cognitive flexibility and interpretability.
• To enable active recognition, inference, and scheduling of uncertain tasks based on EEG data.

Main Methods:

• A Pseudo Task based Neural State Encoder (PNSE) encodes EEG windows into a structured hyperspherical embedding space.
• A Neural Transition Graph Network (NTGN) learns relationships between cognitive states and tasks.
• A Temporal Pseudo-Task Boundary Model (TPBM) captures the temporal evolution of cognitive states, integrated with a neurosymbolic decision layer and fuzzy inference engine.

Main Results:

• The framework achieved high performance on a subject-independent EEG dataset (COG-BCI).
• Key metrics included Silhouette Score (73.7%), Hit Rate (71.43%), NDCG (91.58%), and MRR (76.67%).
• Precision (81.1%), Recall (83.4%), Accuracy (83.47%), and F1 score (82.7%) demonstrated system effectiveness.

Conclusions:

• The proposed neurosymbolic framework enables cognitive modeling from EEG data for active state recognition and task inference.
• It offers a temporally unified and cognitively flexible foundation for adaptive decision-support systems.
• The model successfully combines the interpretability of symbolic reasoning with the adaptability of neural representation learning.