Summary: Learning Control Algorithms: Convergence and
Robustness
Anatoly A. Pervozvanskii
Department of Mechanics and Control Processes
St.Petersburg State Technical University
St.Petersburg, Russia
e-mail: control1@citadel.stu.neva.ru
Konstantin E. Avrachenkov
Centre for Industrial and Applied Mathematics
School of Mathematics
University of South Australia, The Levels
SA 5095, Australia
e-mail: matkea@zarniwoop.levels.unisa.edu.au
Abstract
The general learning control procedure is considered.
It is shown that known convergence conditions for the
learning procedure demand operators with unproper
transfer functions. Therefore we propose new weaker
convergence conditions which admit learning operators
with proper transfer functions. A new robust learn-

Source: Avrachenkov, Konstantin - INRIA Sophia Antipolis

Collections: Computer Technologies and Information Sciences