Populations of dopamine neurons don't simply signal reward prediction errors--they encode rich maps of possible future outcomes, including when and how much a reward might be. This diverse, adaptable neural code mirrors cutting-edge AI strategies known as distributional reinforcement learning.
Populations of dopamine neurons don't simply signal reward prediction errors--they encode rich maps of possible future outcomes, including when and how much a reward might be. This diverse, adaptable neural code mirrors cutting-edge AI strategies known as distributional reinforcement learning.
Share this Article
Recommended