Reproduce “Implementation of a Model of the Cortex Basal Ganglia Loop” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Naoya Arakawa
Naoya Arakawa of The Whole Brain Architecture Initiative developed a computational model of the cortex-basal ganglia-thalamus (CBT) loop, proposing that the cortex predicts actions while the basal ganglia uses reinforcement learning to gate their execution. The model successfully learned a delayed reward task, demonstrating increasing average rewards and a reduction in cortical prediction error.
Key insights
- The cortex and basal ganglia engage in a synergistic relationship where the cortex refines its action predictions based on actions actually executed and reinforced by the basal ganglia.
- The basal ganglia's primary function is to provide a "Go/NoGo" signal, controlling the timing and initiation of actions proposed by the cortex, rather than selecting the action type itself.
- A simplified, modular computational model can effectively reconcile and integrate predictive learning in the cortex with reinforcement learning in the basal ganglia.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.