Convergence of the Gradient Method in Normed Linear Spaces.

Abstract

Recently Golomb and Tapia defined the gradient of a (C sup 1) functional on an arbitrary normed linear space. They showed that their gradient method coincides with the known method of steepest descent and established Curry's theorem (any cluster point of the gradient sequence is a stationary point) under the assumption that the gradient operator be single valued and continuous. In this paper the authors prove Curry's theorem in the full generality of normed linear spaces. This includes many important applications in which the gradient operator is neither continuous nor single valued.

Document Details

Document Type
Technical Report
Publication Date
Oct 01, 1974
Accession Number
ADA001653

Entities

People

  • R. A. Tapia
  • R. H. Byrd

Organizations

  • University of Wisconsin–Madison

Tags

DTIC Thesaurus Topics

  • Convergence
  • Sequences
  • Stationary

Readers

  • Graph Algorithms and Convex Optimization.
  • Operations Research

Technology Areas

  • Space