Neural networks based approach for computing eigenvectors and eigenvalues of symmetric matrix (Q1767833)

The authors present a new approach to the well-studied problem of computing eigenvectors corresponding to the largest or smallest eigenvalues of a real symmetric matrix. A model of a neural network is proposed, which is described by a certain nonlinear differential equation, involving the given symmetric matrix. An interesting representation of the solutions of the network is provided, together with convergence theorems. Some good numerical examples are given to enlighten the theoretical results. At the present the computational performance is questionable, but it is expected to improve it by implementing the network in the hardware in the future.