Title: Machine learning magnetic parameters from spin configurations

Abstract: Estimation of Hamiltonian is key to our understanding of everything in quantum mechanics. For magnetic systems, in particular, the task can be daunting due to the exponentially large Hilbert space. Here, we show that modern machine learning architectures, particularly the artificial neural networks(ANN), can estimate several parameters of condensed-matter Hamiltonians from merely a configuration. Trained with the properly labeled data, neural networks can achieve the estimation well. We demonstrate the success of the estimation on the magnetic system, where the neural networks successfully estimate the Heisenberg exchange interaction($A_{ex}$), Dzyaloshinskii-Moriya interaction($DMI$) and saturation magnetization($M_{sat}$) out of the spin configuration on both simulation and experimental observation. This methodology provides a paradigm to estimate all possible parameters of Hamiltonian of system.