The 81st JSAP Autumn Meeting, 2020

Presentation information

Oral presentation

FS Focused Session "AI Electronics" » FS.1 Focused Session "AI Electronics"

[9p-Z28-1~19] FS.1 Focused Session "AI Electronics"

Wed. Sep 9, 2020 1:30 PM - 7:00 PM Z28

Tetsuya Asai(Hokkaido Univ.), Nakajima Mitsumasa(NTT)

6:30 PM - 6:45 PM

[9p-Z28-18] Q-Learning based on Chaotic Laser Time Series

〇(M1)Takashi Urushibara1, Nicolas Chauvet1, Satoshi Kochi2, Satoshi Sunada2, Kazutaka Kanno3, Atsushi Uchida3, Makoto Naruse1 (1.Tokyo Univ., 2.Kanazawa Univ., 3.Saitama Univ.)

Keywords:Laser chaos, Q-Learning

Q-Learning is a reinforcement learning method wherein evaluation values for actions are adapted for multiple states. In this study, we regard a Q-learning system as a parallel array of multi-armed bandit problems to search for a better action for each state while using chaotic laser time series. The method exhibits superior performances compared to other methods such as randomly-shuffled chaotic laser time series, uniformly and normally distributed random numbers, indicating that the temporal structure inherent in chaos may provide positive effects.