An Analysis and Optimization Technique for Probabilistic Graphs,
Abstract
A communication network serving multiple user pairs is modeled as a probabilistic graph. The problem of finding optimum constrained strategies for either attacking or designing the network is considered. An objective function which can be interpreted as the expected value of the network is developed. New network analysis algorithms which create the objective function for arbitrary networks are presented. The objective function is nonseparable and, in general, neither concave nor convex; consequently conventional optimization techniques fail. An algorithm which enumerates a subset of the feasible strategies, and evaluates a subset of these, to find the optimum is presented. Two heuristics are given as well. All the algorithms herein have been programmed in Fortran IV. Computational results and execution times for a number of examples are given. Comparisons with results in the literature are given where possible. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1976
- Accession Number
- ADA028398
Entities
People
- William P. Dotson Jr
Organizations
- Air Force Institute of Technology