Remapping in a recurrent neural network model of navigation and context inference

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• Neurons in navigational brain regions encode spatial information and remap firing patterns in response to context changes.
• Understanding how these circuits maintain local computations while adapting to global context shifts is a key challenge.

Purpose of the Study:

• To investigate how navigational circuits can preserve local computations while responding to global context changes.
• To develop a computational model explaining neural remapping in response to environmental and task variations.

Main Methods:

• Trained recurrent neural network models to perform simultaneous navigation (position tracking) and context inference.
• Simulated environments with transiently-cued context changes to observe model behavior.

Main Results:

• The models exhibited activity patterns similar to population-wide remapping observed in the entorhinal cortex.
• The proposed model demonstrated generalization to more complex navigation and inference tasks.

Conclusions:

• A single neural circuit can perform both navigation and context inference.
• Remapping can be explained by a unified computational mechanism for processing spatial information and contextual cues.