[18p-438-1~11] 3.10 Optical quantum physics and technologies

[18p-438-1] Formation of a large-scale optical neural network by using optical multiple reflections and its application to reservoir computing

〇Satoshi Sunada¹, Tomoaki Niiyama¹, Atsushi Uchida², Kenichi Arai³ (1.Kanazawa Univ., 2.Saitama Univ., 3.NTT CS Lab.)

Keywords:reservoir computing, optical neural network, microcavity

Reservoir computing (RC) is a new computational paradigm for temporal information processing, and it has been regarded as a powerful tool for solving highly complex and abstract computational tasks without algorithmic instructions. In this work, we propose a new photonic RC, which enables efficient and fast information processing at a microscale. The information processing is based on an ultrahigh dimensional mapping of input information by optical spatiotemporal wave dynamics, which is created by multiple scattering inside a microcavity. In this talk, I will numerically demonstrate the computational capabilities of the proposed RC for basic benchmark tests.