Supporting material for the article:
By Manuel
López-Ibáñez and Thomas Stützle
March 2010 (last update: July 2010)
Manuel López-Ibáñez and Thomas
Stützle. The
impact of design choices of multi-objective ant colony optimization algorithms
on performance: An experimental study on the biobjective TSP. In
GECCO 2010, pages 71–78. ACM press, New York, NY, 2010.
★ Best paper award
[ bibtex |
doi: 10.1145/1830483.1830494 |
supplementary
information ]
Abstract:
Over the last few years, there have been a number of proposals of ant
colony optimization (ACO) algorithms for tackling multiobjective combinatorial
optimization problems. These proposals adapt ACO concepts in various ways, for
example, some use multiple pheromone matrices and multiple heuristic matrices
and others use multiple ant colonies. In this article, we carefully examine
several of the most prominent of these proposals. In particular, we identify
commonalities among the approaches by recasting the original formulation of
the algorithms in different terms. For example, several proposals described in
terms of multiple colonies can be cast equivalently using a single ant colony,
where ants use different weights for aggregating the pheromone and/or the
heuristic information. We study algorithmic choices for the various proposals
and we identify previously undetected trade-offs in their performance.
Keywords: Multiobjective Optimization, Ant Colony
Optimization, Travelling Salesman Problem
MOAQ: β = 0 vs. β = 2
These plots illustrate the different behaviour between β = 0 vs.
β = 2. Each file contains plots of the output of 15 independent runs with
different random seeds. Plots are given for a limit of iterations of 1000,
5000 and 25000, and for a time limit of n/5, n and
5n seconds, where n is the instance size.
P-ACO: Single vs. multiple heuristic matrices
Each plots displays differences in the EAFs of two alternatives. Each EAF
is calculated from 15 independent runs with different random seeds. Plots are
given for a limit of iterations of 1000, 5000 and 25000, and for a time limit
of n/5, n and 5n seconds, where n is the
instance size.
BicriterionAnt: One vs. three colonies
BicriterionAnt: Weighted product vs. weigthed sum
mACO-1 vs. mACO-2
mACO-3 vs. mACO-4
Comparison of MOACO algorithms vs. BicriterionAnt with 3
colonies