Accession Number : AD0736095
Title : On Boyse's Method for Undiscounted Markov Renewal Programming--An Improved Algorithm and a New Proof.
Descriptive Note : Research rept.,
Corporate Author : CARNEGIE-MELLON UNIV PITTSBURGH PA MANAGEMENT SCIENCES RESEARCH GROUP
Personal Author(s) : Morton,Thomas E.
Report Date : DEC 1971
Pagination or Media Count : 19
Abstract : Recently Boyse has presented yet a third method for extending White's modified successive approximation procedure from Markov decision programming to markov renewal programming, in addition to those proposed by Schweitzer and this author. Although his procedure requires much more computation and storage than the latter methods, it is unique in generalizing the property that finite horizon solutions are provided as intermediate output. The rate of convergence of the finite horizon problem with horizon length is often of great interest to the practitioner who plans to use the infinite horizon stationary result as an approximation to a more realistic non-stationary problem. In the paper a shorter, more insightful derivation is given of convergence and bounds for Boyse's method, and a class of improved algorithms are proposed. (Author)
Descriptors : (*DECISION THEORY, *MATHEMATICAL PROGRAMMING), STOCHASTIC PROCESSES, APPROXIMATION(MATHEMATICS), CONVERGENCE, ALGORITHMS
Subject Categories : Operations Research
Distribution Statement : APPROVED FOR PUBLIC RELEASE