Three types of remapping with linear decoders: a population-geometric perspective

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Neuroscience

Background:

• Hippocampal remapping is crucial for spatial memory, allowing place cells to form distinct environmental maps.
• Existing theories explain remapping through memory interference or latent state shifts, but a unified understanding is lacking.

Purpose of the Study:

• To unify and elucidate the mechanisms underlying hippocampal remapping.
• To provide a framework for understanding diverse remapping theories and experimental findings.

Main Methods:

• Utilizing a neural coding and population geometry perspective.
• Modeling hippocampal population activity within a linearly-decodable latent space.
• Simulating and visualizing three distinct remapping mechanisms.

Main Results:

• Identified three primary mechanisms for remapping: changes in neural-to-latent space mapping, non-spatial mixed selectivity modulation, and redundant coding via neural variability.
• Demonstrated these mechanisms in a network model, relating them to existing literature.
• Provided a unifying framework for interpreting remapping phenomena.

Conclusions:

• Hippocampal remapping can be comprehensively understood through a latent space framework, encompassing distinct coding and variability mechanisms.
• This framework aids in visualizing, understanding, and comparing diverse remapping theories and experimental data.
• The model serves as a testbed for neural response variability in various experimental contexts.