Research activity in 2001
1.Efficiency and efficacy of a symmetric variant of the Simplex method
are in course of evaluation
2.Development of methodologies in learning processes headed by the
computer (AUDIP project)
3.Development, implementation and experimentation of global
optimization algorithms in order to detect the minimal energy
configuration according to the Lennard-Jones potential function as well as
for other relevant potentials such as the Morse potential. The minimal
energy of a cluster of identical and uncharged atoms, where the energy
between any pair of atoms is represented by the Lennard-Jones pair
potential or the Morse pair potential, is searched. This problems are
widely explored in the literature, they are very challenging from the
optimization point of view and have important applications in the field of
chemistry.
Currently, the combination of two-phase local searches, previously
successully employed, with other methods, such as basin-hopping, is tested
with quite encouraging results.
4. Max-clique problems, i.e. finding the maximum clique (the complete
subgraph of a graph with highest cardinality) are very challenging
optimization problems. An analysis of existing methods based on a
continuous formulation of such problem revealed their equivalence with
particular combinatorial heuristics. Such heuristics have been further
developed and lead to the detection of the global optimum or at least of
the best known value for many benchmark problems.
5. Function landscapes have been analyzed in order to better understand
why some algorithms perform particularly well on some test functions.
During the analysis it has also been revealed that a particular test
function (the Griewank function) has the surprisinmg property of becoming
easier as its dimension increases. The reason of such behavior has now
been explained.