Modularity Facilitates Classification Performance of Spiking Neural Networks for Decoding Cortical Spike Trains

Area of Science:

• Computational neuroscience
• Artificial intelligence
• Neural networks

Background:

• Spiking neural networks (SNNs) incorporate recurrence, enhancing classification but not yet surpassing artificial neural networks.
• Modularity, a key feature of biological brains, has not been extensively explored for improving SNN performance.

Purpose of the Study:

• To investigate the impact of modularity on SNN performance.
• To compare the classification accuracy of a modular SNN against a uniform SNN.

Main Methods:

• Proposed a modular SNN architecture.
• Compared the modular SNN with a uniform SNN using cortical spike train classification tasks.

Main Results:

• The modular SNN demonstrated a significant performance improvement over the uniform SNN.
• Performance gains increased with network size and decreased with the number of modules.

Conclusions:

• Modularity can enhance SNN performance, suggesting potential for improved AI and brain-machine interfaces.
• Modular SNNs may serve as valuable models for studying neuronal spike synchrony.