Information processing, dimensionality reduction and reinforcement learning in the basal ganglia

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Basal ganglia models have evolved significantly over 20 years.
• Early models focused on single pathways, later models introduced multiple pathways and movement disorder insights.
• Current models emphasize action selection and sequence generation, but new findings suggest a different approach.

Purpose of the Study:

• To review the evolution of basal ganglia models.
• To introduce and explore the reinforcement driven dimensionality reduction (RDDR) model.
• To highlight the RDDR model's implications for basal ganglia function and disease.

Main Methods:

• Literature review of basal ganglia modeling.
• In-depth analysis of the reinforcement driven dimensionality reduction (RDDR) model.
• Discussion of experimental predictions and computational capacity.

Main Results:

• The RDDR model proposes basal ganglia compress cortical information via reinforcement signals.
• This model offers novel insights into the computational mechanisms of the basal ganglia.
• It provides new experimental predictions for understanding basal ganglia function in health and disease.

Conclusions:

• The RDDR model represents a paradigm shift in understanding basal ganglia information processing.
• This model offers a new framework for investigating basal ganglia roles in neurological disorders.
• Further research is needed to experimentally validate the RDDR model's predictions.