An opponent striatal circuit for distributional reinforcement learning.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Machine Learning

Background:

• Distributional reinforcement learning (RL) enhances machine learning by considering reward distributions.
• The mesolimbic dopamine system is implicated in RL, updating mean value representations in the striatum.
• Neural encoding of higher-order reward moments in the brain remains largely unknown.

Conclusions:

• The striatum encodes reward variance, extending traditional RL models.
• Dopamine modulates distributional representations in the striatum.
• A model of D1/D2 MSN opponency explains distributional RL benefits.