The Effects of Biological Constraints on the Performance of a Spiking Neural Network

カテゴリ: 部門大会

論文No: PS2-2

グループ名: 【C】2022年電気学会電子・情報・システム部門大会

発行日: 2022/08/24

タイトル(英語): The Effects of Biological Constraints on the Performance of a Spiking Neural Network

著者名: Li Bin（東京大学）,井口 龍輝（東京大学）,野山 大樹（東京大学）,Zheng Tianyi（東京大学）,小谷 潔（東京大学）,神保 泰彦（東京大学）

著者名(英語): Bin Li (The University of Tokyo),Ryuki Iguchi (The University of Tokyo),Hiroki Noyama (The University of Tokyo),Tianyi Zheng (The University of Tokyo),Kiyoshi Kotani (The University of Tokyo),Yasuhiko Jimbo (The University of Tokyo)

キーワード: Spiking Neural Network|Transfer Learning

要約(日本語): Spiking Neural Network(SNN) has shown an impressive power efficiency stemming from its high biological plausibility. Although it has been reported that the SNN converted from continues Recurrent Neural Network(RNN) can realize a parallel performance when compared with its counterpart RNN, the most remarkable character within SNN—biological plausibility is lost to a large extent. To apply biological constraints and elucidate the mechanism of the influence would be essential for rebuilding the biological plausibility (Dale’s principle, etc.) and constructing more power-efficient SNN. Here we studied the effects of applying several types of biological constraints on the performance of an SNN converted from continues RNN and proposed an improved SNN with biological constraints.