Structured experience shapes strategy learning and neural dynamics in the medial entorhinal cortex

Area of Science:

• Neuroscience
• Computational Biology
• Animal Behavior

Background:

• Prior experience is crucial for animals to solve complex tasks flexibly.
• The structure of early training experiences, not just their presence, influences learning and generalization.
• Neural mechanisms underlying how early training shapes learning remain largely unknown.

Purpose of the Study:

• To investigate how the structure of early training experiences shapes learning and generalization in recurrent neural networks (RNNs) and mice.
• To identify neural mechanisms by which structured training promotes adaptive behavior.

Main Methods:

• Trained RNNs on variants of an odor-timing task.
• Used behavioral and electrophysiological recordings in mice trained with staged sequences.
• Applied dynamical systems approaches to analyze network activity.

Main Results:

• RNNs and mice trained without structured early experience showed rigid strategies and errors.
• Animals and networks with balanced early training demonstrated better generalization and adaptive neural activity.
• Structured curricula led to distinct dynamical motifs in networks, enabling correct abstractions.

Conclusions:

• The structure of early learning experiences is a critical determinant of behavioral flexibility and generalization.
• Structured training fosters the development of adaptable neural representations and computational mechanisms.
• Findings highlight the importance of curriculum design for promoting generalizable knowledge in both biological and artificial systems.