Integrable discretization and its applications (Q1807793)

Exploiting integrable discretization techniques and numerical methods for solving ordinary differential equations, the author develops new algorithms for computing the largest eigenvalue of real symmetric matrices. This is done by formulating the calculation of Rayleigh quotient as a solution of a gradient system and by studying the exact solution of this system. Thereafter, applying Euler and Runge-Kutta methods to an equivalent system of linear differential equations yields new numerical algorithms and recovers the inverse iteration, power and Wilkinson's methods. Comparison of the new algorithm and the Wilkinson's method and numerical examples are discussed.