A study on multiform multi-objective evolutionary optimization

Multi-objective optimization problem (MOP) denotes the optimization problem involving more than one objective function to be optimized simultaneously. In the literature, to solve MOP, evolutionary algorithm has been recognized as an effective approach. Over the years, a number of multi-objective evolutionary algorithms (MOEAs) have been developed. In this paper, we present a study on multiform multi-objective evolutionary optimization. In contrast to existing MOEAs, which only focus on the optimization of a single MOP, the proposed new paradigm considers to construct multiple forms of a given MOP, which may contain different useful information for solving the MOP. The evolutionary search is then performed on both the given MOP and the constructed forms concurrently. By transferring useful traits found along the evolutionary search across the given MOP and the built problem forms, enhanced multi-objective optimization performance can be obtained. To the best of our knowledge, there is no existing work that considers the multiform optimization for solving MOP. To evaluate the performance of the proposed multiform paradigm for multi-objective optimization, comprehensive empirical studies with commonly used MOP benchmarks using different existing MOEAs as the basic MOP solvers are conducted and analyzed.