Presentation Information
[10p-N106-8]Gradient-based optimization of spin-torque oscillator
〇(P)Yusuke Imai1, Shuhong Liu1, Nozomi Akashi2, Kohei Nakajima1 (1.Univ. of Tokyo, 2.Kyoto Univ.)
Keywords:
reservoir computing,spintronics
We propose a method for optimizing the parameters (magnetic field, current, and damping constant) of the LLG equation, which describes the dynamics of a spin torque oscillator, using the gradient descent method. We then apply this method to parameter estimation tasks and image recognition tasks (MNIST, CIFAR-10) to demonstrate that tuning the parameters of the LLG equation improves accuracy.