Summary

International Symposium on Nonlinear Theory and Its Applications

2022

Session Number:B1L-E

Session:

Number:B1L-E-01

Long-Tailed Distribution of Excitatory Postsynaptic Potentials Enhances Learning Performance of Liquid State Machine

Ibuki Matsumoto ,   Sou Nobukawa ,   Nobuhiko Wagatsuma ,   Tomoki Kurikawa,  

pp.244-247

Publication Date:12/12/2022

Online ISSN:2188-5079

DOI:10.34385/proc.71.B1L-E-01

PDF download (353.7KB)

Summary:
In the cerebral cortex, excitatory postsynaptic potentials (EPSPs) exhibit a long-tailed distribution. It is known that the long-tailed characteristic of EPSPs induces spontaneous activity and stochastic resonance. In this context, we hypothesized that a long-tailed distribution of EPSPs would improve the learning performance of machine learning. Therefore, we input a spiking neural network with long-tailed characteristics of EPSPs into liquid state machine (LSM), and then evaluated the learning performance of LSM via a memory capacity task. These results suggest that long-tailed distributions of EPSPs enhance higher memory capacity. This finding might help improve the learning performance of LSM.