Incorporating Cultural Adaptation and Local Search Intometa-heuristics Continuous Optimisation

Solving challenging problems with non-separable, multi-funnel, multi-property, asymmetrical characteristics are the targets of recent studies in continuous optimisation. In this thesis, I proposed a local search-hybrid optimisation framework with a view to solving some problems belonging to the mentioned class. I tried to achieve this goal by following three approaches. Firstly, to utilise the strengths of both population-based and single-point based optimisation methods, in the framework I proposed a population of trajectory individuals, where each individual performs a single point based method. Secondly, I focused on combining some successful search strategies of existing meta-heuristics algorithms with some newly proposed ideas. Thirdly, to ensure that the suitable search strategy is automatically chosen when solving a certain problem, I proposed a method allowing individuals exchange their knowledge with each other to adjust their behaviours adaptively. To implement this, I adopted the cultural adaptation paradigm from Cultural Algorithms. The proposed optimisation framework, called Cultural Algorithms –Iterated Local Search (CA-ILS), was tested in various test functions. The results show that averagely CA-ILS perform better than its predecessors: CA and ILS. CA-ILS also belongs to the top three when compared with top algorithms from the CEC’05 Special Session in the tested functions.