Presentation Information

[9a-N106-10]Evaluation of Learning Performance in m-VSTO Reservoirs with Sequential Dual-Inputs

〇Kota Horizumi1, Takahiro Chiba2, Takashi Komine1 (1.Ibaraki Univ., 2.Yamagata Univ.)

Keywords:

Physical reservoir computing,spin torque oscillator

Vortex spin-torque oscillators (VSTOs) are promising candidates for physical reservoir devices. Introducing an additional layer, the modified VSTO (m-VSTO) enables switching between periodic and chaotic dynamics simply by adjusting the drive current. In this study, we numerically investigate whether reservoir-learning performance can be enhanced by combining outputs obtained at two current levels that correspond to distinct dynamical regimes.