著者

中嶋 浩平

出版者

一般社団法人 システム制御情報学会

雑誌

システム/制御/情報 (ISSN:09161600)

巻号頁・発行日

vol.63, no.12, pp.505-511, 2019-12-15 (Released:2020-06-15)

参考文献数

57

著者

中嶋 浩平

出版者

一般社団法人 人工知能学会

雑誌

人工知能 (ISSN:21882266)

巻号頁・発行日

vol.33, no.5, pp.570-576, 2018-09-01 (Released:2020-09-29)

著者

明石 望洋 國吉 康夫 城 健智 西田 三博 櫻井 良 若尾 康通 中嶋 浩平

雑誌

2023年第70回応用物理学会春季学術講演会

巻号頁・発行日

2023-01-24

著者

中嶋 浩平

出版者

公益社団法人 応用物理学会

雑誌

応用物理学会学術講演会講演予稿集 第80回応用物理学会秋季学術講演会 (ISSN:24367613)

巻号頁・発行日

pp.248, 2019-09-04 (Released:2022-07-22)

著者

窪田 智之 中嶋 浩平 高橋 宏知

出版者

一般社団法人 電気学会

雑誌

電気学会論文誌C(電子・情報・システム部門誌) (ISSN:03854221)

巻号頁・発行日

vol.140, no.7, pp.723-729, 2020-07-01 (Released:2020-07-01)

参考文献数

29

---

Neuronal systems are dynamical. In the dynamical system, the externally-driven responses, called transient dynamics, are stimulus-specific but reproducible. Under the assumption that the neuronal system is deterministic, we here hypothesized that such reproducible transient activities could produce computational capability in a chaotic dynamical system. To test this hypothesis, we estimated the maximal Lyapunov exponent of neuronal activities in the primary visual cortex (V1) of mice, and quantified their information processing capacity in the transient dynamics. Consequently, V1 was characterized as a chaotic system, where almost identical input time-series led to different trajectories. We also demonstrated that, when mice were visually stimulated with drifting gratings, the trajectories contained the input time-series information for at least 5.2 s after stimulation. These results suggest that computational capability in V1 emerges from reproducible transient activities in the chaotic system. Yet, the estimate information processing capacities in V1 were much lower than those in theoretical studies. Further verification is still required to elucidate the discrepancy between theoretical and experimental results.

著者

中嶋 浩平

出版者

横断型基幹科学技術研究団体連合(横幹連合)

雑誌

横幹連合コンファレンス予稿集 第8回横幹連合コンファレンス

巻号頁・発行日

pp.B-1-3, 2017 (Released:2018-02-18)

---

Reservoir computing (RC) was first proposed as a framework to train recurrent neural networks. In this framework, a low-dimensional input is projected to high-dimensional dynamical systems, which are typi-cally referred to as a reservoir. If the dynamics of the reservoir involve adequate nonlinearity and memory, em-ulating nonlinear dynamical systems only requires adding a linear, static readout from the high-dimensional state space of the reservoir. Because of its generic nature, RC is not limited to digital simulations of neural net-works, and any high-dimensional dynamical system can serve as a reservoir if it has the appropriate properties. The approach using a physical entity rather than abstract computational units as a reservoir is called physical reservoir computing (PRC). In this presentation, several novel platforms based on PRC are introduced using physical substrates. These platforms include soft materials (e.g., silicone-based soft robotic arm) and faraday waves generated on the water surface (which we call, “physical liquid state machines”), and they illustrate the potentials of the framework through a number of experiments. The focus will particularly be on how dynamical system aspects can provide a novel view of the PRC framework, including the relevance of noise-induced phe-nomena or random dynamical systems.