Using mixed mode programming to parallelize an indicator-based evolutionary algorithm for inferring multiobjective phylogenetic histories

doi:10.1007/s00500-016-2219-6

Return

Using mixed mode programming to parallelize an indicator-based evolutionary algorithm for inferring multiobjective phylogenetic histories

2016-06-14

PRE

Sergio Santander‐Jiménez *

Miguel A. Vega‐Rodríguez

DOI：10.1007/s00500-016-2219-6

Original

Original request for help

Save

Abstract

En 中文

Multiple problems in bioinformatics research involve the optimization of time-consuming objective functions over exponentially growing search spaces. The capabilities shown by modern parallel systems composed of clusteredmulticoremultiprocessors represent an opportunity to address such difficult problems. A suitable paradigm to exploit these systems lies on the combination of mixed mode programming and evolutionary computation. This research focuses on the reconstruction of multiobjective phylogenetic hypotheses by using an indicator-based evolutionary algorithm. In order to overcome the main sources of complexity of the problem, we propose a parallel adaptation of this algorithm based on master-worker principles. Experimental results on six real data sets report that the design achieves an efficient exploitation of a shared-distributed memory hybrid system composed of 48 processing cores, observing improved scalability in comparison with other parallel proposals. In addition, the inferred Pareto fronts give account of the relevance of the indicator-based design, verifying significant solution quality under different multiobjective metrics and biological testing procedures.

Keywords:

Parallel computing

Multiobjective optimization

Bioinformatics

Phylogenetic inference

AI Summary

Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.