On the Convergence of Asynchronous Parallel Pattern Search

Kolda, Tamara Gilbson

Sandia National Laboratories

关键词: Nonlinear Problems;    Optimization;    Convergence;    Pattern Recognition;    99 General And Miscellaneous//Mathematics, Computing, And Information Science;   

DOI  :  10.2172/793723
RP-ID  :  SAND2001-8696
RP-ID  :  AC04-94AL85000
RP-ID  :  793723

美国|英语

来源: UNT Digital Library

【 摘 要 】

In this paper the authors prove global convergence for asynchronous parallel pattern search. In standard pattern search, decisions regarding the update of the iterate and the step-length control parameter are synchronized implicitly across all search directions. They lose this feature in asynchronous parallel pattern search since the search along each direction proceeds semi-autonomously. By bounding the value of the step-length control parameter after any step that produces decrease along a single search direction, they can prove that all the processes share a common accumulation point and that such a point is a stationary point of the standard nonlinear unconstrained optimization problem.