A New Estimation Distributed Algorithm Applied to a Many-Objective Discrete Optimization Problem

Abstract

Many-Objective Optimization Problems are problems that have more than three objective functions. For a small number of objective functions, Multi-Objective Evolutionary Algorithms provide good results, but when the number of objective functions grows, these algorithms present scalability problems. In this paper we focus on Multi-Objective Discrete Problems (MODO) with many objectives. We propose a new Estimation Distributed Algorithm (EDA) applied to MODO, called ArchEDA, with the aim of improving the results achived by MOEAs. The main idea is to combine EDA with archiving methods, in order to select the solutions used on the probabilistic models. To evaluate Arch-EDA we apply the algorithm to the 0/1 Multi-Objective Knapsack Problem, considering two to ten objective functions and a set of benchmarking instances. The results achieved are compared, through statistical analysis, with the NSGA-III, NSGA-II and SPEA2 algorithms.