Self-organizing continuous attractor network models of hippocampal spatial view cells.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• Hippocampal spatial view cells encode primate visual environment locations.
• These neurons maintain firing without visual input, suggesting internal memory mechanisms.
• Continuous attractor networks are standard models for neural networks with spatial memory.

Purpose of the Study:

• To model primate hippocampal spatial view cells that maintain spatial firing without visual input.
• To investigate how idiothetic (self-motion) inputs update spatial representations in the dark.
• To explore different ways continuous attractor networks integrate velocity signals.

Main Methods:

• Developed three computational models of hippocampal spatial view cells.
• Models utilize continuous attractor networks to integrate spatial information.
• Incorporated velocity signals from head and eye movements, and positional information from head direction and eye position cells.

Main Results:

• The models demonstrate how spatial view cells maintain firing and update representations using idiothetic cues in absence of vision.
• Two models integrate head and eye velocity signals using a 'memory trace' learning rule.
• A third model uses head direction and eye position cells to update spatial views in the dark.

Conclusions:

• Continuous attractor networks can model primate spatial view cells that retain and update spatial information using self-motion cues.
• The models provide insights into the neural mechanisms underlying spatial navigation and memory.
• Different integration strategies for idiothetic signals are possible within attractor network architectures.