Simon Martin

A multi-agent based cooperative approach to scheduling and routing

In this paper, we propose a general agent-based distributed framework where each agent is implementing a different metaheuristic/local search combination. Moreover, an agent continuously adapts itself during the search process using a direct cooperation protocol based on reinforcement learning and pattern matching. Good patterns that make up improving solutions are identified and shared by the agents. This agent-based system aims to provide a modular flexible framework to deal with a variety of different problem domains. We have evaluated the performance of this approach using the proposed framework which embodies a set of well known metaheuristics with different configurations as agents on two problem domains, Permutation Flow-shop Scheduling and Capacitated Vehicle Routing. The results show the success of the approach yielding three new best known results of the Capacitated Vehicle Routing benchmarks tested, whilst the results for Permutation Flow-shop Scheduling are commensurate with the best known values for all the benchmarks tested.

The 'composite' design pattern in metaheuristics

Introduced in [13], the ‘Composite’ Design Pattern can be observed whenever some ‘aggregator’ object can be considered to share behavioural commonalities with the entities from which it is composed. It therefore offers a concise means of describing functional [9] or structural recursion [5], which are of course ubiquitous in computer science and software engineering. In computer science, the prototypical example is a recursive data structure such as a tree, defined inductively either as an empty node or else as a node having a list of nodes as its children. In software engineering, an example is a concept such as Displayable, instances of which might include OKButton, (which can directly display itself), or Window which might display itself by asking all the Displayables it contains to display themselves in turn. Figure 1 shows the class diagram for a prototypical example of the composite pattern [13]. It is useful to be able to identify Composite relationships within metaheuristics for several reasons:

A Re-characterization of Hyper-Heuristics

Hyper-heuristics are an optimization methodology which ‘search the space of heuristics’ rather than directly searching the space of the underlying candidate-solution representation. Hyper-heuristic search has traditionally been divided into two layers: a lower problem-domain layer (where domain-specific heuristics are applied) and an upper hyper-heuristic layer, where heuristics are selected or generated. The interface between the two layers is commonly termed the “domain barrier”. Historically this interface has been defined to be highly restrictive, in the belief that this is required for generality. We argue that this prevailing conception of domain barrier is so limiting as to defeat the original motivation for hyper-heuristics. We show how it is possible to make use of domain knowledge without loss of generality and describe generalized hyper-heuristics which can incorporate arbitrary domain knowledge.

BBIOS: a characterization of evolutionary algorithm stability

We report preliminary results of linking distinct parameter metrics and stability via a novel dynamical system stability characterization (BBIOS). We conduct EA trials to determine the extent of EA stability in parameter space neighborhoods defined by metrics. We capture EA performance loss due to perturbation.