[18a-A302-1~8] 3.10 Optical quantum physics and technologies

[18a-A302-2] Photonic Reinforcement Learning by Time-division Multiplexing of Laser Chaos

〇Makoto Naruse¹, Takatomo Mihana², Hirokazu Hori³, Hayato Saigo⁴, Kazuya Okamura⁵, Atsushi Uchida² (1.NICT, 2.Saitama Univ.,, 3.Univ. Yamanashi, 4.Nagahama Inst., 5.Nagoya Univ.)

Keywords:reinforcement learning, decision making, laser chaos

Reinforcement learning is an accurate decision making in an uncertain environment and forms the foundation of many important applications. In our former study, we successfully solved the multi-armed bandit problem by utilizing laser chaos. However, the scalability towards practical applications was unresolved. In this research, we show a scalable principle by time-division multiplexing the sampling of laser chaos and demonstrate experimental results.