A Simple Distributed Random Bit Climber on Many-Objective Epistatic Problems
Yudai TAGAWA, Hernan E. Aguirre, Kiyoshi Tanaka · IEICE Transactions on Information and Systems · 2026
We study a distributed bit climber algorithm for many-objective optimization of binary problems. This algorithm decomposes the many-objective problem into a minimum number of single-objective problems, specified by the original evaluation functions and one additional scalarizing function that computes the solution hypervolume. Random bit climbers optimize separately the single-objective functions until they reach a local optimum and restart their search from a bounded population of non-dominated solutions collected from the solutions generated by all climbers. In this work, we study the climbing behavior of dRBC evaluating the method on subclasses of epistatic problems using MNK-landscapes and compare with moRBC, MOEA/D and, NSGAIII, showing that the simpler distributed bit climber is superior to the other MOEAs.