Dynamic system for solving complex eigenvalue problems

In the paper, the authors propose a dynamic system for solving complex eigenvalue problems. We show that, under some mild conditions, the set of the unit eigenvectors corresponding to the eigenvalue with the largest real part is asymptotically stable and the sets of the unit eigenvectors corresponding to the other eigenvalues is unstable. The proposed dynamical system can be realised as analogue neural networks for real-time applications.