Functional network reorganization in motor cortex can be explained by reward-modulated Hebbian learning

Area of Science:

• Neuroscience
• Computational Neuroscience
• Machine Learning

Background:

• Neuroprosthetic device control relies on motor cortex neuron adaptation.
• Previous studies showed selective tuning changes in monkey motor cortex neurons to correct for misinterpretations.
• Understanding these self-tuning properties is crucial for improving brain-computer interfaces.

Purpose of the Study:

• To propose and validate a simple learning rule explaining the self-tuning properties of motor cortex neurons.
• To demonstrate that this rule can account for experimentally observed neuronal adaptation in neuroprosthetic control.
• To investigate the role of neuronal noise and reward in this learning process.

Main Methods:

• Development of a computational model based on a novel reward-modulated Hebbian learning rule.
• The rule utilizes neuronal noise for exploration and global reward signals for weight updates.
• Model simulations were performed with parameters fitted to experimental data from monkey motor cortex recordings.

Main Results:

• The proposed learning rule successfully explains the selective adaptation of neuronal tuning properties.
• The model demonstrates efficient performance optimization within biologically realistic timescales, even under high noise levels.
• Simulated learning effects closely replicate those observed in monkey experiments when neuronal noise levels are matched.

Conclusions:

• A simple, biologically plausible learning rule can explain adaptive neuroprosthetic control.
• The rule's ability to leverage neuronal noise without explicit signal differentiation is a key feature.
• This work provides a theoretical framework for understanding and enhancing learning in brain-computer interfaces.