Abstract
Fejer processes are frequently used models for many iterative algorithms in optimization and related areas. They can be combined with different kinds of decomposition schemes and generate various projection-type methods suitable for parallel computations. This paper reviews some recent results on Fejer processes with diminishing disturbances and suggests a new adaptive parameter-free stepsize control rule for such algorithms.
Acknowledgements
The author wishes to thank G.T. Herman and two anonymous referees for a great amount of meticulous reading of preliminary versions of this paper and many constructive suggestions.
The study was supported by RFBR grant 09-01-90413-Ukr_f_a.