A multiplicative reinforcement learning model capturing learning dynamics and interindividual variability in mice.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Animal Behavior

Background:

• Individual differences in learning speed are common in humans and animals.
• Standard learning models attribute this variability to differences in learning rates.
• The precise sources of interindividual variability in learning remain unclear.

Purpose of the Study:

• To investigate the role of initial neural connectivity in determining learning speed.
• To test the hypothesis that initial synaptic weights influence learning dynamics.
• To compare learning performance using multiplicative versus additive learning rules.

Main Methods:

• Theoretical modeling of neural learning processes.
• Experimental training of mice on an auditory Go/NoGo discrimination task and subsequent reversal.
• Analysis of learning curves using computational models.

Main Results:

• Mice exhibited variable learning curves with initial delays during task acquisition.
• Reversal learning was consistently faster than initial learning across all subjects.
• A reinforcement learning model with a multiplicative rule accurately fitted learning curves and explained variability.
• An additive learning rule failed to capture the observed learning dynamics.

Conclusions:

• Multiplicative learning rules are crucial for understanding biological learning dynamics.
• Initial synaptic connectivity is a significant determinant of learning speed and interindividual differences.
• Brain's initial wiring may predispose individuals to learn certain tasks faster.