LisBON: A framework for parallelisation and hybridisation of optimisation algorithms
Christoph Dürr, Tim Fühner, Ponnuthurai Nagaratnam Suganthan · 2007
This paper introduces LisBON, a novel framework for distributed, hybrid optimisation algorithms. LisBON aims at simplifying the development of memetic algorithms-a combination of heuristic, population-based search approaches with local optimisers. Moreover, LisBON’s design allows for an integration of virtually any optimisation algorithm. It could hence be used to implement a large variety of different hybrid approaches, multiple-restart methods in local search routines, and multiple populations and meta-evolution in evolutionary algorithms. With LisBON, it is not only possible to distribute optimisers onto different computing nodes, but also the concurrent evaluation of merit functions can be defined in a straight forward manner. In this paper, we present the design of LisBON and its key components. Furthermore, as an example, the steps required to develop a memetic algorithm are explained. It is shown that the obtained hybrid method is able to outperform the underlying genetic algorithm in terms of convergence speed on an established benchmark function (Griewangk).