Prediction of magnetization dynamics in a reduced dimensional feature space setting utilizing a low-rank kernel method

We establish a machine learning model for the prediction of magnetization dynamics as function external field described by Landau-Lifschitz-Gilbert equation, partial differential equation motion in micromagnetism. The allows fast and accurate determination response to an which is illustrated thin-film standard problem. data-driven method internally reduces dimensionality problem means nonlinear reduction unsupervised learning. This not only makes time steps possible, but also decisively complexity process where states from simulated micromagnetic associated with different fields are used input data. use truncated representation kernel principal components describe between predictions. capable handling large training sample sets owing low-rank approximation matrix extension component analysis ridge regression. approach entirely shifts computations into reduced dimensional setting breaking down dimension thousands tens.