Computing quasi-differentials and exhausters by function values

Authors

  • M.Yu. Andramonov

Keywords:

оптимизация
квазидифференциалы
метод Ньютона
метод наискорейшего спуска
квазидифференциальное исчисление
субдифференциалы

Abstract

A general method for computing quasi-differentials and co-differentials is proposed; the method is based on their approximation by polygons. The accuracy of approximation depends on the numbers of vectors for which the directional derivative is computed. The algorithm can be used for nonsmooth optimization, in particular, in Newton’s method and the steepest descent method for minimizing complicated non-differentiable functions.

New computational technologies

Downloads

Published

2006-07-03

Issue

Vol. 7 (2006): Issue 2.

Section

Section 1. Numerical methods and applications

Author Biography

M.Yu. Andramonov

Kazan Federal University,
Chebotarev Institute of Mathematics and Mechanics

References

  1. Демьянов В.Ф. Теорема о неподвижной точке в негладком анализе и ее применение. СПб: Изд-во Санкт- Петербургского ун-та, 1996.愦灭;percent - - - 134с.
  2. Demyanov V.F. Exhausters and convexificators: new tools in nonsmooth analysis. Quasi-differentiability and related topics. Dordrecht: Kluwer Academic Publishers, 2001.
  3. Andramonov M.Yu., Rubinov A.M., Glover B.M. Cutting angle methods in global optimization // Applied Mathematics Letters. May 1999.
  4. Balas E. Disjunctive programming // Annals of Discrete Mathematics. 1979. 5. 3-51.
  5. Beaumont F. Algorithm for disjunctive programming problems // European Journal of Operational Research. 1990. 42, N 3. 362-371.
  6. Demyanov V.F., Rubinov A.M. Constructive non-smooth analysis. Frankfurt: Peter Lang, 1995.
  7. Grossmann I., Lee S. New algorithms for generalized nonlinear disjunctive programming // Computers and Chemical Engineering. 2000. 24. 2125-2141.
  8. Rubinov A.M. Abstract convexity and global optimization. Dordrecht: Kluwer Academic Publishers, 2000.