A spike timing dependent plasticity learned feedback network in reservoir computing for timeseries pattern recognition

カテゴリ：国際会議

論文No：D4

グループ名：【C】AVIC2025

発行日：2025/10/20

著者名：Akinobu Yamaguchi (Nihon University), Yoshiki Sasaki (Nihon University)

キーワード：Reservoir computing,VLSI,Spiking neural network,Spike timing dependent plasticity

要約(英語)：In recent years, research into engineering applications of the biological brain's superior characteristics, such as its parallel processing capabilities, low power consumption, and compact implementation—has been gaining significant attention. Among these, Reservoir Computing (RC), which excels at time-series processing, has become a key focus. In our previous work, we constructed the reservoir layer of an RC system using the P-HCNM, a spiking neuron model built with analog electronic circuits used in Spiking Neural Networks (SNNs). However, to utilize the information generated by the reservoir layer, a readout layer must be designed. Applying the backpropagation algorithm, widely used in software-based neural networks, to SNNs is difficult. Therefore, in this study, we propose a feedback network Configuration for applying Spike Timing Dependent Plasticity (STDP)to the readout layer. We investigated whether both excitatory and inhibitory learning could be achieved based on the difference in output frequencies between the reservoir and output layers. The results demonstrate that the proposed network is capable of both excitatory and inhibitory learning depending on these frequency differences.

本誌掲載ページ：81-84p

原稿種別：英語

PDFファイルサイズ：528Kバイト