Never-Ending Learning for Explainable Brain Computing

Area of Science:

• Cognitive Neuroscience
• Artificial Intelligence
• Neuroscience

Background:

• Human intelligence and behavior exploration is complex due to fragmented findings in cognitive neuroscience.
• A unified and transparent interpretation of diverse study results remains a significant challenge.

Purpose of the Study:

• To propose an explainable brain computing framework for continuous cognitive neuroscience investigation.
• To integrate evidence combination and fusion computing within a Knowledge-Information-Data (KID) architecture.
• To enhance the understanding of brain activity patterns related to human intelligence.

Main Methods:

• Employs a never-ending learning paradigm with a Knowledge-Information-Data (KID) architecture.
• Utilizes joint knowledge-driven forward inference and data-driven reverse inference.
• Incorporates internal evidence learning (multi-task neuroimaging) and external evidence learning (topic modeling of studies), with human-in-the-loop mechanisms.

Main Results:

• Reveals intricate uncertainties in human reasoning brain localization through two case studies.
• Demonstrates the framework's capability for continuous brain cognition investigation.
• Highlights the potential of systematization to advance explainable brain computing.

Conclusions:

• The proposed framework offers a path toward more systematic and explainable brain computing.
• Advancements in systematization can lead to a finer-grained understanding of neural correlates of human intelligence.
• The integration of multi-modal evidence and human interaction is crucial for robust cognitive modeling.