Category: Business Analytics and Intelligence
Copyright © 2014, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
304
Biased Randomization of
Classical Heuristics
INTRODUCTION
In the context of combinatorial optimization
problems, this chapter discusses how to random-
ize classical heuristics in order to transform these
deterministic procedures into more efficient proba-
bilistic algorithms. This randomization process
can be performed by using a uniform probability
distribution or, even more interesting, by using a
non-symmetric distribution.
Combinatorial Optimization Problems (COPs)
have posed numerous challenges to the human
mind throughout the past decades. From a theo-
retical perspective, they have a well-structured
definition consisting of an objective function
that needs to be minimized or maximized, and a
series of constraints that must be satisfied. From
a theoretical point of view, these problems have
an interest on their own due to the mathematics
involved in their modeling, analysis and solution.
However, the main reason for which they have
been so actively investigated is the tremendous
amount of real-life applications that can be suc-
cessfully modeled as a COP. Thus, for example,
decision-making processes in fields such as lo-
gistics, transportation, and manufacturing contain
plentiful hard challenges that can be expressed as
COPs (Faulin et al., 2012; Montoya et al., 2011).
Accordingly, researchers from different areas
–e.g. Applied Mathematics, Operations Research,
Computer Science, and Artificial Intelligence–
have directed their efforts to conceive techniques
to model, analyze, and solve COPs.
A considerable number of methods and al-
gorithms for searching optimal or near-optimal
solutions inside the solution space have been
developed. In some small-sized problems, the solu-
tion space can be exhaustively explored. For those
instances, efficient exact methods can usually
provide the optimal solution in a reasonable time.
Unfortunately, the solution space in most COPs is
exponentially astronomical. Thus, in medium- or
large-size problems, the solution space is too large
and finding the optimum in a reasonable amount of
time is not a feasible task. An exhaustive method
that checks every single candidate in the solution
space would be of very little help in these cases,
since it would take exponential time. Therefore, a
large amount of heuristics and metaheuristics have
been developed in order to obtain near-optimal
solutions, in reasonable computing times, for
medium- and large-size problems, some of them
even considering realistic constraints.
Angel A. Juan
IN3-Open University of Catalonia, Spain
José Cáceres-Cruz
IN3-Open University of Catalonia, Spain
Sergio González-Martín
IN3-Open University of Catalonia, Spain
Daniel Riera
IN3-Open University of Catalonia, Spain
Barry B. Barrios
IN3-Open University of Catalonia, Spain
DOI: 10.4018/978-1-4666-5202-6.ch028