Derivation of effective low energy models using machine learning

We introduce a machine learning protocol to extract an effective low-energy spin model from a Kondo Lattice Model (KLM) with classical localized moments. The resulting effective spin model reproduces the phase diagram obtained with the original KLM and uncovers the effective four-spin interactions that are responsible for the stability of the skyrmion crystal phase. It enables an efficient computation of static and dynamical properties that are numerically orders of magnitude faster than the original KLM. Even though information about the spin dynamics is not used as a part of the training dataset, comparison of dynamical spin structure factor in the fully polarized phase reveals a reasonable agreement for the magnon dispersion.