Inferring Multiobjective Phylogenetic Hypotheses by Using a Parallel Indicator-Based Evolutionary Algorithm

The application of multiobjective optimization techniques to solve biological problems has significantly grown in the last years. In order to generate satisfying approximations to the Pareto-optimal set, two key problems must be addressed. Firstly, we must distinguish solution quality in accordance with the optimization goal, usually measured by means of multiobjective quality indicators. Secondly, we must undertake the development of parallel designs to carry out searches over exponentially growing solution spaces. This work tackles the reconstruction of phylogenetic relationships by applying an Indicator-Based Evolutionary Algorithm. For this purpose, we propose a parallel design based on OpenMP which considers the computation of hypervolume-based indicators in fitness assignment procedures. Experiments on four biological data sets show significant results in terms of parallel scalability and multiobjective performance with regard to other methods from the literature.