We demonstrate the performance of magnetic nanowire lattices for reservoir computing using micromagnetic simulation. Domain walls are injected into nanowires as the input and are driven by clocking magnetic fields. The static magnetization states of wires are used as the reservoir states. The distribution of node weights shows that the device saves the memory as a shift register, and the domain walls provide the nonlinear computational power for reservoir computing. The maximum short-term memory capacity of 5 and the parity-check capacity of 5 are achieved. We expect the maximum capacities can be boosted by increasing the number of asymmetric structures and the length of the wires.
Numerical simulation of reservoir computing with magnetic nanowire lattices without inversion symmetry
Note: This paper is part of the APL Special Collection on Neuromorphic Computing: From Quantum Materials to Emergent Connectivity.
K. Hon, K. Takahashi, K. Enju, M. Goto, Y. Suzuki, H. Nomura; Numerical simulation of reservoir computing with magnetic nanowire lattices without inversion symmetry. Appl. Phys. Lett. 10 January 2022; 120 (2): 022404. https://doi.org/10.1063/5.0073465
Download citation file: