[14a-A501-1~13] 10.2 Fundamental and exploratory device technologies for spin

▼ [14a-A501-3] Simulation of reservoir computing using dipole-coupled nanomagnet array with different clocking frequency

〇(M1)Kazuki Tsujimoto¹, Hikaru Nomura^1,2, Yuki Kuwabiraki¹, Ryoichi Nakatani¹, Yoshishige Suzuki^1,2 (1.Osaka Univ., 2.CSRN-Osaka)

Keywords:reservoir computing, nanomagnet, neural network

Neural network has made remarkable achievements in recent years. However, in general AI, higher performance requires more energy consumption. To reduce the energy consumption, reservoir computing using physical phenomenon for calculation has been reported. We have demonstrated simulation of reservoir computing using the stable magnetization state of dipole-coupled nanomagnet array. Besides, some group reported reservoir computing using spin torque oscillator. Therefore magnetization dynamic state of dipole-coupled nanomagnet array can be also useful for reservoir computing.
In this presentation, we demonstrate simulation of reservoir computing using dipole-coupled nanomagnet array with different clocking frequency. We set clocking frequency from hundred kHz order to GHz order, and investigate short-term memory capacity and parity check capacity of the reservoir.