Recently, photonic implementation of artificial neural networks (ANNs) has catching interests because they have a great potential to reduce the operational power and latency beyond the electronic computing. The photonic circuit can solve large scale matrix operation, which is a dominant factor of ANN computation, with ultrafast propagation speed thanks to their inherent parallelism in space, frequency and time division. Here, I demonstrate a reservoir computing — a randomly connected recurrent neural network — specified on-chip photonic circuit capable of operating at sub-Peta-scale Multiply-Accumulate per second speeds [1]. I also explain the relationship between deep neural network and wave equation in the waveguide, which enables large-scale integration of photonic neuromorphic circuit in future [2]. [1] M. Nakajima et al., Commun. Phys. 4, 20 (2021). [2] M. Nakajima et al., arXiv:2006.13541

Mitsumasa Nakajima received the M.E. and Ph.D. degrees in material science from the Tokyo Institute of Technology, Tokyo, Japan, in 2010 and 2015, respectively. In 2010, he joined Nippon Telegraph and Telephone (NTT) Laboratories, where he was involved in the development of large-scale optical switches. Recent his research interests are optical switches and their applications including neuromorphic photonics and optical signal processing for telecom. He was a recipient of the 8 research award including Young Engineer Award from the Institute of Electronics, Information and Communication Engineers (IEICE) of Japan