Catálogo de publicaciones - libros

This paper presents a simple way of combining inference with stochastic search for solving constraint satisfaction problems. The approach makes use of an evolutionary algorithm for search assisted by an inference algorithm, the variable elimination procedure. The hybrid algorithm obtained is adapted in such way that a balance between exploitation and exploration is preserved. The results are presented for the Max-CSP optimization task.

The social golfer problem (SGP) has attracted significant attention in recent years because of its highly symmetrical, constrained, and combinatorial nature. Nowadays, it constitutes one of the standard benchmarks in the area of constraint programming. This paper presents the first evolutionary approach to the SGP. We propose a memetic algorithm (MA) that combines ideas from evolutionary programming and tabu search. In order to lessen the influence of the high number of symmetries present in the problem, the MA does not make use of recombination operators. The search is thus propelled by selection, mutation, and local search. In connection with the latter, we analyze the effect of baldwinian and lamarckian learning in the performance of the MA. An experimental study shows that the MA is capable of improving results reported in the literature, and supports the superiority of lamarckian strategies in this problem.

The graph set T -colouring problem (GSTCP) is a generalisation of the classical graph colouring problem and it is used to model, for example, the assignment of frequencies in mobile networks. The GSTCP asks for the assignment of sets of nonnegative integers to the vertices of a graph so that constraints on the separation of any two numbers assigned to a single vertex or to adjacent vertices are satisfied and some objective function is optimised. Among the various objective functions of interest, we focus on the minimisation of the span, that is, the difference between the largest and the smallest integers used. In practical applications large size instances of the GSTCP are to be solved and heuristic algorithms become necessary. In this article, we propose a new hybrid procedure for the solution of the GSTCP that combines a known tabu search algorithm with an algorithm for the enumeration of all feasible re-assignments of colours to a vertex. We compare the new algorithm with the basic tabu search algorithm and for both we study possible variants. The experimental comparison, supported by statistical analysis, establishes that the new hybrid algorithm performs better on a variety of instance classes.

It is general knowledge that hybrid approaches can improve the performance of search heurististics. The first phase, exploration, should detect regions of good solutions, whereas the second phase, exploitation, shall tune these solutions locally. Therefore a combination (hybridization) of global and local optimization techniques is recommended. Although plausible at the first sight, it remains unclear how to implement the hybridization, e.g., to distribute the resources, i.e., number of function evaluations or CPU time, to the global and local search optimization algorithm. This budget allocation becomes important if the available resources are very limited. We present an approach to analyze hybridization in this case. An evolution strategy and a quasi-Newton method are combined and tested on standard test functions.