Improving a local search technique for network optimization using inexact forecasts

Authors:
Gilberto Flores Lucio;Martin J. Reed;Ian D. Henning
Affiliations:
University of Essex, UK;University of Essex, UK;University of Essex, UK
Venue:
ICN'05 Proceedings of the 4th international conference on Networking - Volume Part I
Year:
2005

Citing 11
Cited 0

Fast approximation algorithms for multicommodity flow problems

STOC '91 Proceedings of the twenty-third annual ACM symposium on Theory of computing
Finding a feasible course schedule using Tabu search

Discrete Applied Mathematics - Special issue: Timetabling and chromatic scheduling
Telecommunications network design algorithms

Telecommunications network design algorithms
Tabu search

Modern heuristic techniques for combinatorial problems
Genetic algorithms + data structures = evolution programs (2nd, extended ed.)

Genetic algorithms + data structures = evolution programs (2nd, extended ed.)
Guided Genetic Algorithm and its Application to Radio Link Frequency Assignment Problems

Constraints
A simple local-control approximation algorithm for multicommodity flow

SFCS '93 Proceedings of the 1993 IEEE 34th Annual Foundations of Computer Science
Evolutionary programming techniques for constrained optimizationproblems

IEEE Transactions on Evolutionary Computation
Local search genetic algorithm for optimal design of reliablenetworks

IEEE Transactions on Evolutionary Computation
Applying an evolutionary algorithm to telecommunication networkdesign

IEEE Transactions on Evolutionary Computation
Fast local search and guided local search and their application to British Telecom's workforce scheduling problem

Operations Research Letters

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents an evolutionary computation approach to optimise the design of communication networks where traffic forecasts are uncertain. The work utilises Fast Local Search (FLS), which is an improved hill climbing method and uses Guided Local Search (GLS) to escape from local optima and to distribute the effort throughout the solution space. The only parameter that needs to be tuned in GLS is called the regularization parameter lambda (λ). This parameter represents the degree up to which constraints on the features in the optimization problem are going to affect the outcome of the local search. To fine-tune this parameter, a series of evaluations were performed in several network scenarios to investigate the application towards network planning. Two types of performance criteria were evaluated: computation time and overall cost. Previous work by the authors has introduced the technique without fully investigating the sensitivity of λ on the performance. The significant result from this work is to show that the computational performance is relatively insensitive to the value of λ and a good value for the problem type specified is given.