Rapid learning of predictive maps with STDP and theta phase precession.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• The predictive map hypothesis suggests the hippocampus predicts future states.
• The successor representation (SR) formalizes this, with place cells encoding expected future occupancy.
• Current models rely on temporal difference learning, which lacks known hippocampal implementation.

Purpose of the Study:

• To investigate a biologically plausible mechanism for learning the successor representation in the hippocampus.
• To demonstrate how spike-timing dependent plasticity (STDP) can approximate the SR.
• To explain experimentally observed hippocampal phenomena using this novel learning mechanism.

Main Methods:

• Developed a computational model using spiking neurons and theta sweeps.
• Implemented spike-timing dependent plasticity (STDP) for learning.
• Simulated temporally compressed trajectories ('theta sweeps') to approximate SR.

Main Results:

• STDP on theta sweeps rapidly learns an approximation of the successor representation.
• The model explains phenomena like backward expansion and place field elongation.
• It also accounts for topographical ordering of place field sizes along the dorsal-ventral axis.

Conclusions:

• Hebbian learning via STDP on theta sweeps provides a biologically plausible mechanism for hippocampal SR learning.
• This model successfully explains key aspects of hippocampal function and spatial representation.
• The findings offer insights into the neural basis of predictive coding and spatial memory.