In this paper, we present a general approach for solving constraint problems by local search. The proposed approach is based on a set of high-level constraint primitives motivated by constraint programming systems. These constraints constitute the basic bricks to formulate a given combinatorial problem. A tabu search engine ensures the resolution of the problem such formulated. Experimental results are shown to validate the proposed approach.

Graph coloring is a well known problem from graph theory that, when solving it with local search algorithms, is typically treated as a series of constraint satisfaction problems: for a given number of colors k, one has to find a feasible coloring; once such a coloring is found, the number of colors is decreased and the local search starts again. Here we explore the application of Iterated Local Search to the graph coloring problem. Iterated Local Search is a simple and powerful metaheuristic that has shown very good results for a variety of optimization problems. In our research we investigate different perturbation schemes and present computational results on some hard instances from the DIMACS benchmark suite.

Code optimization and high level synthesis can be posed as constraint satisfaction and optimization problems, such as graph coloring used in register allocation. Naturallyoccurring instances of such problems are often small and can be solved optimally. A recent wave of improvements in algorithms for Boolean satisfiability (SAT) and 0-1 ILP suggests generic problem-reduction methods, rather than problem-specific heuristics, because (1) heuristics are easily upset by new constraints, (2) heuristics tend to ignore structure, and (3) many relevant problems are provably inapproximable. The NP-spec project offers a language to specify NP-problems and automatic reductions to SAT. Problem reductions often lead to highly symmetric SAT instances, and symmetries are known to slow down SAT solvers. In this work, we compare several avenues for symmetry-breaking, in particular when certain kinds of symmetry are present in all generated instances. Our surprising conclusion is that instance-independent symmetries should often be processed together with instance-specific symmetries rather than earlier, at the specification level. 1

This paper presents a new neighbourhood structure for Graph Colouring problems based upon permutations of the colours of vertices in a subgraph. This neighbourhood is an exponentially large subspace of the solution space, for which a local optimum can be found in polynomial time using an algorithm called Permutation Descent. We thus provide

Abstract. Graph coloring is a well known problem from graph theory that, when attacldng it with local search algorithms, is typically treated as a series of constraint satisfaction problems: for a given number of colors k one has to find a feasible coloring: once such a coloring is found, the number of colors is decreased and the local search starts again. Here we explore the application of Iterated Local Search on the graph coloring problem. Iterated Local Search is a simple and powerful metaheuristic that has shown very good results for a variety of optimization problems. In our research we investigated several perturbation schemes and present computational results on a widely used set of benchmarks problems, a sub-set of those available from the DIMACS benchmark suite. Our results suggest that Iterated Local Search is particularly promising on hard, structured graphs.

The graph set T-colouring problem (GSTCP) generalises the classical graph colouring problem; it asks for the assignment of sets of integers to the vertices of a graph such 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 objective functions of interest is the minimisation of the difference between the largest and the smallest integers used (the span). In this article, we present an experimental study of local search algorithms for solving general and large size instances of the GSTCP. We compare the performance of previously known as well as new algorithms covering both simple construction heuristics and elaborated stochastic local search algorithms. We investigate systematically different models and search strategies in the algorithms and determine the best choices for different types of instance. The study is an example of design of effective local search for constraint optimisation problems.

In this paper, we present a first scatter search approach for the Graph Coloring Problem (GCP). The evolutionary strategy scatter search operates on a set of configurations by combining two or more elements. New configurations are improved before replacing others according to their quality (fitness), and sometimes, to their diversity. Scatter search has been applied recently to some combinatorial optimization problems with promising results. Nevertheless, it seems that no attempt of scatter search has been published for the GCP. This paper presents such an investigation and reports experimental results on some wellstudied DIMACS graphs.

this paper. In particular they used a neighborhood operator and iterative improvement not used in our adaptation of MOGLS

We introduce two novel evolutionary formulations of the problem of coloring the nodes of a graph. The first formulation is based on the relationship that exists between a graph's chromatic number and its acyclic orientations. It views such orientations as individuals and evolves them with the aid of evolutionary operators that are very heavily based on the structure of the graph and its acyclic orientations. The second formulation, unlike the first one, does not tackle one graph at a time, but rather aims at evolving a "program" to color all graphs belonging to a class whose members all have the same number of nodes and other common attributes. The heuristics that result from these formulations have been tested on some of the Second DIMACS Implementation Challenge benchmark graphs, and have been found to be competitive when compared to the several other heuristics that have also been tested on those graphs.

We present a search space analysis and its application in improving local search algorithms for the graph coloring problem. Using a classical distance measure between colorings, we introduce the following clustering hypothesis: the high quality solutions are not randomly scattered in the search space, but rather grouped in clusters within spheres of specific diameter. We first provide intuitive evidence for this hypothesis by presenting a projection of a large set of local minima in the 3D space. An experimental confirmation is also presented: we introduce two algorithms that exploit the hypothesis by guiding an underlying Tabu Search (TS) process. The first algorithm (TS-Div) uses a learning process to guide the basic TS process toward as-yet-unvisited spheres. The second algorithm (TS-Int) makes deep investigations within a bounded region by organizing it as a tree-like structure of connected spheres. We experimentally demonstrate that if such a region contains a global optimum, TS-Int does not fail in eventually finding it. This pair of algorithms significantly outperforms the underlying basic TS algorithm; it can even improve some of the best-known solutions ever reported in the literature (e.g. for dsjc1000.9). Key words: graph coloring, local optima distribution, search by learning. 1