Multi-View Broad Learning System for Primate Oculomotor Decision Decoding.

Area of Science:

• Neuroscience
• Machine Learning
• Computational Neuroscience

Background:

• Multi-view learning leverages diverse data sources to enhance model performance.
• Cortical neural signals, including local field potentials (LFPs) and action potentials (spikes), offer complementary information for brain state decoding.
• Primate brain state decoding is crucial for understanding cognitive processes.

Purpose of the Study:

• To extend the Broad Learning System (BLS) to a multi-view learning framework.
• To decode monkeys' oculomotor decisions using simultaneously recorded medial frontal LFPs and spikes.
• To evaluate the effectiveness of the proposed multi-view BLS against existing methods.

Main Methods:

• Developed a multi-view extension of the Broad Learning System (BLS).
• Utilized simultaneously recorded local field potentials (LFPs) and action potentials (spikes) from non-human primates.
• Applied the multi-view BLS to decode oculomotor decisions.

Main Results:

• Medial frontal LFPs and spikes contain complementary information relevant to oculomotor decisions.
• The multi-view BLS significantly outperformed classical and state-of-the-art single-view and multi-view learning approaches.
• Demonstrated the efficacy of integrating multi-view neural data for enhanced decoding accuracy.

Conclusions:

• Multi-view learning, particularly with the extended BLS, is a powerful approach for decoding primate brain states.
• The complementary nature of LFPs and spikes can be effectively harnessed for improved neural decoding.
• This work advances the application of machine learning in neuroscience for understanding decision-making processes.