Abstract
The behaviour of the standard steepest-descent algorithm for a quadratic function in IR d is investigated. By rescaling the iterates to remain always on the unit sphere one can reveal special features of this behaviour: the renormalized algorithm converges to a two-point cycle on the unit circle; the cycle depends on the starting point in a complicated manner (the set of points converging to the same cycle is fractal), but all cycles belong to a particular plane, given by certain eigenvectors of the Hessian matrix of the objective function. The stability of the attractor is analysed. The rate of convergence of the algorithm is investigated. It is shown that the worst value of this rate is obtained only for some particular starting points. The introduction of a relaxation coefficient in the steepest-descent algorithm completely changes its behaviour, which may become chaotic. Different attractors are presented. We show that relaxation allows a significantly improved rate of convergence.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Akaike, H. (1959). On a successive transformation of probability distribution and its application to the analysis of the optimum gradient method. Ann. Inst. Statist. Math. Tokyo, 11:1–16.
Forsythe, G. (1968). On the asymptotic directions of the s-dimensional optimum gradient method. Numerische Mathematik, 11:57–76.
Hale, J. and Koçak, H. (1991). Dynamics and Bifurcations. Springer-Verlag, Heidelberg.
Luenberger, D. (1973). Introduction to Linear and Nonlinear Programming. Addison-Wesley, Reading, Massachusetts.
Pronzato, L., Wynn, H., and Zhigljaysky, A. (2000). Dynamical Search. Chapman & Hall/CRC, London et al.
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2001 Kluwer Academic Publishers
About this chapter
Cite this chapter
Pronzato, L., Wynn, H.P., Zhigljavsky, A.A. (2001). Some Convergence Properties of the Steepest Descent Algorithm Revealed by Renormalisation. In: Hadjisavvas, N., Pardalos, P.M. (eds) Advances in Convex Analysis and Global Optimization. Nonconvex Optimization and Its Applications, vol 54. Springer, Boston, MA. https://doi.org/10.1007/978-1-4613-0279-7_29
Download citation
DOI: https://doi.org/10.1007/978-1-4613-0279-7_29
Publisher Name: Springer, Boston, MA
Print ISBN: 978-0-7923-6942-4
Online ISBN: 978-1-4613-0279-7
eBook Packages: Springer Book Archive