Numerical simulation of artificial spin ice for reservoir computing
Kwan Hon, Yuki Kuwabiraki, Minori Goto, Ryoichi Nakatani, Yoshishige Suzuki, Hikaru Nomura
Abstract
Abstract We propose the use of artificial spin ice (ASI) for reservoir computing and evaluate its performance from simulation results. Our ASI reservoir is formed by 72 magnets arranged in a honeycomb lattice and the temperature is 0 K. A pseudo-random binary sequence is sent to the reservoir and its status is updated by external magnetic fields. Short-term memory capacity of 3.5 and nonlinear computational capacity of 2.9 are achieved when the strength of the magnetic fields is near the switching field of the magnets. The performance can be optimized by tuning the aspect ratios of the magnets.
Topics & Concepts
Spin iceNonlinear systemMagnetMagnetic fieldBinary numberLattice (music)Reservoir computingSequence (biology)Computer simulationMaterials scienceStatistical physicsComputer scienceGeologyMechanicsPhysicsMechanical engineeringMathematicsEngineeringArtificial neural networkArtificial intelligenceChemistryAcousticsBiochemistryMagnetic monopoleQuantum mechanicsRecurrent neural networkArithmeticNeural Networks and Reservoir ComputingQuantum many-body systemsPhysics of Superconductivity and Magnetism