A learning optimization algorithm in graph theory: versatile search for extremal graphs using a learning algorithm

Authors:
Gilles Caporossi;Pierre Hansen
Affiliations:
GERAD and HEC Montréal, Canada;GERAD and HEC Montréal, Canada
Venue:
LION'12 Proceedings of the 6th international conference on Learning and Intelligent Optimization
Year:
2012

Citing 4
Cited 0

Variable neighborhood search

Computers and Operations Research
Variable neighborhood search for extremal graphs: 1 the AutoGraphiX system

Discrete Mathematics
Finding relations in polynomial time

IJCAI'99 Proceedings of the 16th international joint conference on Artificial intelligence - Volume 2
Brief Communication: Graphs with maximum connectivity index

Computational Biology and Chemistry

Quantified Score

Hi-index	0.00

Visualization

Abstract

Using a heuristic optimization module based upon Variable Neighborhood Search (VNS), the system AutoGraphiX's main feature is to find extremal or near extremal graphs, i.e., graphs that minimize or maximize an invariant. From the so obtained graphs, conjectures are found either automatically or interactively. Most of the features of the system relies on the optimization that must be efficient but the variety of problems handled by the system makes the tuning of the optimizer difficult to achieve. We propose a learning algorithm that is trained during the optimization of the problem and provides better results than all the algorithms previously used for that purpose.