Adaptive Systems

Informational inequalities in gradient stochastic optimization optimal feasible algorithms

A. V. Nazin

Moscow

Abstract: It is required to find the minimum point of a function, unknown in advance, from measurements of its gradient with a random additive noise. Lowerbounds are obtained of the accuracy of solving this problem in the class of «any» algorithms. Algorithms are demonstrated and analyzed for which these bounds are achieved.

UDC: 517.977.58


Received: 23.10.1987

 English version: Automation and Remote Control, 1989, 50:4, 531–540

Bibliographic databases:

• 
•