ニューロエンジニアリング Neuroengineering

P3-2-253 位相縮約法によるシリコンニューロンのダイナミカルシステムデザイン Dynamical Systems Design of Silicon Neurons using Phase Reduction Method ○中田一紀1, 三浦佳二2, 浅井哲也3 ○Kazuki Nakada1, Keiji Miura2, Tetsuya Asai3 九州大学 稲盛フロンティア研究センター1, 東北大学 大学院情報科学研究科2, 北海道大学 大学院情報科学研究科3 Advanced Electronics Research Divisin, INAMORI Frontier Research Center, Kyushu University, Fukuoka, Japan1, Collaborative Mathematics Research Unit, Graduate School of Information Sciences, Tohoku University, Sendai, Japan2, Graduate School of Information Science and Technology, Hokkaido University, Sapporo, Japan3 Dynamical systems design approaches based the mathematical structure of nonlinear systems have been recently developed in the field of neuromorphic engineering. In the dynamical systems design approaches, the mathematical structure of nonlinear devices and circuits are analyzed as a dynamical system, and the qualitative nature of the devices and circuits are implemented to reproduce desirable dynamical behaviors on a practical physical platform. In terms of the mathematical structures embedded in the devices and circuits, the dynamical systems approaches can be classified at least three: (i) the phase plane and nullcline-based design, (ii) the potential-based design, and (iii) the phase response curve-based design. From the viewpoint of the phase reduction theory, we propose to apply the third approach to designing silicon neurons. Firstly, we clarify design criteria by analyzing the phase response properties of various silicon neurons presented in previous works. Secondly, we design the resonate-and-fire neuron circuit as a specific silicon neuron according to the design criteria. Finally, we show that the synchronization properties of a fully-connected network of the resonate-and-fire neuron circuits with transmission delays can be improved by tuning the phase response properties of the element circuit.

上部に戻る

前に戻る