Synaptic dynamics: Linear model and adaptation algorithm.

Area of Science:

• Computational Neuroscience
• Artificial Intelligence

Background:

• Spiking neural networks (SNNs) are powerful computational models inspired by biological brains.
• Accurate modeling of synaptic dynamics, including short-term plasticity (STP) and long-term potentiation (LTP), is crucial for SNNs.
• Training and analyzing complex SNNs with realistic synaptic behaviors remain challenging.

Purpose of the Study:

• To introduce a novel linear model for synapse temporal dynamics.
• To present a biologically plausible learning algorithm for synaptic adaptation in SNNs.
• To demonstrate the model's and algorithm's efficacy in simulating and training large-scale neural circuitry.

Main Methods:

• Developed a linear model to accurately capture synaptic facilitation and depression.
• Proposed a learning rule capable of simultaneously adjusting LTP and STP parameters.
• Trained a small-scale SNN using the developed model and learning rule to generate specific neural firing patterns.

Main Results:

• The linear model accurately represents synapse temporal dynamics, simplifying SNN analysis and training.
• The learning algorithm effectively adjusts synaptic parameters, mimicking biological learning.
• Simulations demonstrated the system's capability to generate diverse cortical neuron spike and bursting patterns.

Conclusions:

• The proposed linear model and STDP-based learning algorithm offer a practical approach for simulating and training large-scale SNNs.
• This framework facilitates the study of complex neural circuitry with significant numbers of neurons and synapses.
• The method enhances the efficiency and biological plausibility of SNN research.