Passive exposure to task-relevant stimuli enhances categorization learning.

Area of Science:

• Neuroscience
• Machine Learning
• Cognitive Science

Background:

• Perceptual decision-making is typically learned through effortful practice and feedback.
• The role of passive, feedback-free stimulus exposure in learning remains underexplored.

Purpose of the Study:

• To investigate how passive stimulus exposure influences active learning in perceptual decision tasks.
• To compare the efficacy of different training schedules combining passive and active learning phases.
• To model the underlying mechanisms using neural networks.

Main Methods:

• Mice were trained on a sound-categorization task using varied schedules of passive exposure and active training.
• Neural network models with diverse architectures and learning rules were trained on the same task.
• Behavioral data and network performance were analyzed to assess learning efficiency and mechanisms.

Main Results:

• Mice receiving passive exposure, whether before or interleaved with active training, learned faster.
• Neural networks employing unsupervised learning during passive exposure to enhance data separability best explained the behavioral findings.
• Interleaved training showed increased alignment between passive and active learning updates, making few sessions highly effective.

Conclusions:

• Passive exposure is a potent enhancer of active learning in perceptual tasks.
• Unsupervised learning during passive exposure optimizes neural representations for efficient active learning.
• Optimized training schedules can effectively combine passive and active learning for natural and artificial systems.