Temporal sequence compression by an integrate-and-fire model of hippocampal area CA3

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• Hippocampal place cells encode spatial location, forming sequences during navigation.
• Spatial navigation is conceptualized as a sequence learning problem for the hippocampus.
• Previous research indicates the hippocampus can recall place cell sequences faster than experienced, a phenomenon termed temporal compression.

Purpose of the Study:

• To investigate the neural mechanisms underlying temporal compression in hippocampal sequence recall.
• To demonstrate that a simplified computational model can reproduce temporal compression.

Main Methods:

• Utilized a computational model of hippocampal area CA3.
• Employed integrate-and-fire neurons and unsupervised Hebbian learning.
• Simulated learning and recall phases, manipulating network activity and inhibition.

Main Results:

• The model successfully reproduced temporal compression, recalling place cell sequences at an accelerated rate.
• The degree of compression correlated with activity levels during recall and the associativity timespan during learning.
• Temporal compression resulted from a shift in network dynamics from external input pacing during learning to intrinsic properties during recall.

Conclusions:

• Network dynamics and activity levels are critical for temporal compression in hippocampal sequence recall.
• Lowering inhibition during recall enhances network state transitions, leading to faster sequence recall.
• The extent of temporal compression is constrained by the temporal associations established during the initial learning phase.