Magnetohydrodynamic eigenfunction classification with a Neural Network

We present a Fourier-decomposition-based approach aided by a Neural Network for the classification of the eigenfunctions of an operator appearing in ideal magnetohydrodynamics. The Neural Network is trained on individual Fourier modes, which enhances the robustness of the classification. In our tests, the algorithm correctly classified 93.5% of the data and returned the remaining 6.5% for manual classification. The probability of misidentifying the eigenfunctions is estimated as 0.03%. The discussion is kept quite general allowing for potential applications in other fields.