SEFRON: A New Spiking Neuron Model With Time-Varying Synaptic Efficacy Function for Pattern Classification.

Area of Science:

• Computational Neuroscience
• Machine Learning
• Artificial Neural Networks

Background:

• Biological neurons exhibit adaptable synaptic efficacy, crucial for learning and information processing.
• Existing spiking neural network (SNN) models often lack the dynamic synaptic plasticity observed in biological systems.

Purpose of the Study:

• To introduce a novel spiking neuron model, SEFRON, incorporating a time-varying synaptic efficacy function.
• To develop a supervised learning rule for SEFRON tailored for pattern classification tasks.
• To evaluate SEFRON's performance against established SNNs.

Main Methods:

• The SEFRON model utilizes a sum of amplitude-modulated Gaussian functions to represent time-varying synaptic efficacy.
• A supervised learning rule adjusts synaptic weight amplitudes by minimizing the error between desired and actual postsynaptic firing times.
• SEFRON's efficacy is demonstrated on binary pattern classification and benchmark datasets from the UCI repository.

Main Results:

• SEFRON successfully performs pattern classification, showcasing its computational capabilities.
• A single SEFRON classifier achieves generalization performance comparable to multi-layer, multi-neuron SNNs.
• The model's time-varying synapse allows for continuous switching between excitatory and inhibitory postsynaptic potentials.

Conclusions:

• SEFRON offers an efficient and effective approach to pattern classification using spiking neural networks.
• The model's biologically inspired dynamic synaptic plasticity contributes to its strong performance.
• SEFRON presents a promising alternative to complex SNN architectures for classification problems.