A model for complex sequence learning and reproduction in neural populations.

Area of Science:

• Computational neuroscience
• Systems neuroscience
• Neural plasticity

Background:

• Repetitive temporal activity patterns in vertebrate and mammalian brains are experimentally observed.
• These patterns are crucial for functions like movement, speech, and rhythm generation.
• Existing models for sequence learning face challenges in stability, robustness, and biological plausibility.

Purpose of the Study:

• To present two computational models for learning and reproducing external activity sequences.
• To investigate sequence learning mechanisms within neuronal populations.
• To explore implications for working memory, oscillations, and rhythm generation.

Main Methods:

• Developed two computational models based on densely interconnected excitatory neuron populations with population-level plasticity.
• The first model employs temporally asymmetric Hebbian plasticity to form excitation pathways between populations.
• The second model features two layers with Hebbian plasticity, associating input sequences with ongoing activity for recall.

Main Results:

• The first model exhibits oscillatory behavior through the interplay of excitatory and inhibitory populations, aligning with neocortical findings.
• The second model demonstrates the ability to associate input sequences with network activity, enabling recall without external input.
• Both models show potential for learning and reproducing temporal sequences.

Conclusions:

• The proposed models offer a biologically plausible framework for sequence learning and reproduction.
• These models provide insights into the neural mechanisms underlying temporal pattern generation and working memory.
• The findings suggest a role for population-level plasticity in generating complex brain dynamics.