Abstract:
Neural networks are considered as a new universal approach to the mathematical model construction of systems with distributed parameters. They allow us to find efficiently approximate solutions of initial and boundary value problems for partial differential equations and to take into consideration phenomena and coefficient perturbations.