[16a-A33-1~11] FS.1 Focused Session "AI Electronics"

10:45 AM - 11:00 AM JST(1:45 AM - 2:00 AM UTC)

[16a-A33-7] MNIST classification accuracies by quantum reservoir across multiple NISQ devices

〇Ryuji Sakai¹, Hideaki Oba¹, Hideyuki Nakagawa¹, Kazuki Uematsu¹, Tomoyuki Takeguchi¹, Yutaro Iiyama², Lento Nagano², Ryu Sawada², Junichi Tanaka², Koji Terashi² (1.Toshiba Corp., 2.Tokyo Univ. ICEPP)

Keywords:

Quantum Computer,NISQ,Reservoir

Using a quantum reservoir that does not require parameter learning of quantum circuits and is noise resilient, we compared and evaluated the performance of MNIST classification across multiple NISQ devices. Even on noisy NISQ devices, the accuracy rate was 92.75%, achieving performance close to that of the simulator. On the other hand, when different NISQ devices were used for training and testing, the accuracy rate dropped significantly. This is likely due to differences in noise, especially in read-out errors between devices.