Spiking Neural Models of Neurons and Networks for Perception, Learning, Cognition, and Navigation: A Review

Area of Science:

• Neuroscience
• Computational Neuroscience
• Artificial Intelligence

Background:

• Neural network models are crucial for understanding brain function.
• Existing models include rate-based and spiking neural networks.
• Bridging these models is essential for comprehensive brain process analysis.

Purpose of the Study:

• To review and synthesize the history of neural network models.
• To explain the conversion process between rate-based and spiking neural network models.
• To demonstrate the relevance of this conversion to various brain processes.

Main Methods:

• Review of theoretical and experimental results on neural models.
• Analysis of shunting laws in neurophysiology.
• Hypothesis on spatial patterns as functional units of brain processes (short-term memory and long-term memory traces).

Main Results:

• All rate-based neural network models can be converted into spiking neural network models without loss of explanatory power.
• Conversion often results in gains in explanatory power.
• Spatial patterns learned by spiking neurons exhibit spatiotemporal self-similarity, preserving information during brain development.

Conclusions:

• The conversion between rate-based and spiking neural network models offers significant explanatory power for brain processes.
• Short-term memory and long-term memory traces, represented by spatial patterns, are fundamental to brain function.
• Spatiotemporal self-similarity in spiking neurons is vital for the development of neural networks throughout life.