PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Ants and graph coloring
John Shawe-Taylor and Janez Žerovnik
preprint 2004.

This is the latest version of this eprint.

Abstract

Ants algorithm is an evolutionary search method for solving combinatorial optimization problems. In this note we propose a version of the algorithm for coloring a graph with a fixed number of colors. Based on approximate colorings obtained by single ants, edges of maximal evidence are added so that the new ants generation does only attempt to construct colorings with additional restrictions. Probabilities for addition of misleading edges are shown to be very low. Experimentaly we obtain speedup over not cooperating ants and a long single ant run.

Postscript - Requires a viewer, such as GhostView
EPrint Type:Article
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Theory & Algorithms
ID Code:934
Deposited By:Janez Žerovnik
Deposited On:02 February 2005

Available Versions of this Item