Dopamine encodes deep network teaching signals for individual learning trajectories.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Behavioral Neuroscience

Background:

• Striatal dopamine is crucial for decision-making and learning.
• Individual learning trajectories show significant diversity, posing a challenge to understanding dopaminergic mechanisms.

Purpose of the Study:

• To longitudinally investigate the role of dorsal striatal dopamine in mice learning a decision task.
• To elucidate the mechanisms underlying diverse and systematic individual learning trajectories.

Main Methods:

• Longitudinal measurement and optogenetic manipulation of dorsal striatal dopamine in mice.
• Analysis of strategy transitions and stimulus-choice associations.
• Development and analysis of a deep neural network model with heterogeneous teaching signals.

Main Results:

• Mouse learning trajectories exhibited diverse strategy sequences but systematic transitions.
• Dopamine signals encoded stimulus-choice associations, reflecting strategy transitions.
• Optogenetic manipulation of these associations produced distinct learning effects.

Conclusions:

• Individual learning diversity and systematicity can be explained by heterogeneous teaching signals influencing specific association weights.
• This study reveals biological and mathematical principles governing long-term individual learning trajectories.