Abstract:
|
This paper deals with the Minimum Linear Arrangement problem
from an
experimental point of view. Using a test-suite of sparse
graphs, we
experimentally compare several algorithms to obtain upper
and lower bounds for
this problem. The algorithms considered include Successive
Augmentation
heuristics, Local Search heuristics and Spectral Sequencing.
The test-suite is
based on two random models and ``real life'' graphs. As a
consequence of this
study, two main conclusions can be drawn: On one hand, the
best approximations
are usually obtained using Simulated Annealing, which
involves a large amount
of computation time. However, solutions found with Spectral
Sequencing are
close to the ones found with Simulated Annealing and can be
obtained in
significantly less time. On the other hand, we notice that
there exists a big
gap between the best obtained upper bounds and the best
obtained lower bounds.
These two facts together show that, in practice, finding
lower and upper bounds
for the Minimum Linear Arrangement problem is hard. |