Сorrespondence

Application of Random Noise to Optimization and Recognition Problems

R. Z. Khas'minskii


Abstract: It is generally known that by combining gradient methods of optimization with random search procedures it is possible to find the absolute extremum of a function. In this paper, we examine such an optimization model and give some estimates which prove the convergence to the absolute maximum within the framework of this model. The proposed method may be suitable for recognition problems.

UDC: 621.391.18

Received: 21.04.1965

 English version: Problems of Information Transmission, 1965, 1:3, 89–93