[23p-E102-1~14] FS.1 Focused Session "AI Electronics"

[23p-E102-8] Improving performance by complementary and cooperative learning with ionic liquid reservoirs

〇Asahi Arai¹, Sang-Gyu Koh¹, Taiki Koide¹, Renta Sato¹, Ichiro Ohira¹, Kentaro Kinoshita¹ (1.Tokyo Univ of Sci.)

Keywords:Parallel reservoir, Relaxation time constant, Ionic liquid

Focusing on the temporal similarity with the ion diffusion dynamics involved in the generation of biological signals, we worked on the IL physical reservoir study using the dielectric relaxation of ionic liquids (IL). Using the relationship between the relaxation time constant (τ) and the IL viscosity revealed by the complex impedance method, we confirmed the improvement of the learning accuracy by appropriately matching the time scale of the input signal with τ.In the day's session, we will discuss complementary and cooperative learning, in which reservoirs with different τ are used in parallel.