Abstract
As we mentioned in Chapter 1, the problems of learning in non-stationary situations were rarely a subject of studies, even in the parametric case. Historically the first papers on this topic where occasionally published in the sixties and seventies. The proper tool for solving such a type of problems seemed to be the dynamic stochastic approximation technique [76], [77] as an extension of Robbins-Monro [201] and Kiefer-Wolfowitz [135] procedures for the non-stationary case. The traditional procedure of stochastic approximation was also used [290], [304] with a good effect for tracking the changing regression function root. The dynamic stochastic approximation procedure was also used for the construction of algorithms tracking time-varying moments of random variables with non-stationary probability distributions [53]. Some other results concerning learning in a time-varying environment are scattered in literature (see e.g. [111], [134], [202]).
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 2004 Springer-Verlag Berlin Heidelberg
About this chapter
Cite this chapter
Rutkowski, L. (2004). General Learning Procedure in a Time-Varying Environment. In: New Soft Computing Techniques for System Modeling, Pattern Classification and Image Processing. Studies in Fuzziness and Soft Computing, vol 143. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-40046-2_4
Download citation
DOI: https://doi.org/10.1007/978-3-540-40046-2_4
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-05820-2
Online ISBN: 978-3-540-40046-2
eBook Packages: Springer Book Archive