Multi-objective Random Bit Climbers with Weighted Permutation on Large Scale Binary MNK-Landscapes

Multi-Objective Evolutionary Algorithms have proven to be very effective when solving Multi-Objective Optimization Problems. However, their performance decreases significantly when solving large scale problems, which can have hundreds or thousands of variables. Although several algorithms have been proposed to tackle this problem in the recent years, most of them are designed for continuous problems, and only a few focus on binary ones. In this paper, we propose a modification to multi-objective random one-bit climbers that achieves better performance in large scale binary problems by learning the trend of the values of the decision variables from previously found solutions and applying that information to decide which ones to focus on when executing the bit climb. We present the implemented algorithm, compare its performance to other well known evolutionary algorithms and study some of its properties.