Extending Guided Local Search – Towards a Metaheuristic Algorithm With No Parameters To Tune

Guided Local Search is a general penalty-based optimisation method that sits on top of local search methods to help them escape local optimum. It has been applied to a variety of problems and demonstrated effective. The aim of this paper is not to produce further evidence that Guided Local Search is an effective algorithm, but to present an extension of Guided Local Search that potentially has no parameter to tune. Compared to other algorithms, Guided Local Search is relatively easy to apply, as there is only one major parameter (λ) to set. In some applications, performance of Guided Local Search is insensitive to the value of this parameter. Nevertheless, the value of this parameter can affect the performance of Guided Local Search in some problems. In this paper, we show how (a) an aspiration criterion and (b) random moves may be added to Guided Local Search to reduce the sensitivity of its performance to the parameter value. The extended Guided Local Search is tested on the SAT, weighted MAX-SAT and Quadratic Assignment Problems with positive results.