A Grid-Based Inverted Generational Distance for Multi/Many-Objective Optimization

Assessing the performance of Pareto front (PF) approximations is a key issue in field evolutionary multi/many-objective optimization. Inverted generational distance (IGD) has been widely accepted as indicator for evaluating comprehensive quality PF approximation. However, IGD usually becomes infeasible when facing real-world optimization problem it needs to know true priori. In addition, time complexity grows quadratically with size solution/reference set. To address aforementioned issues, grid-based (Grid-IGD) proposed estimate both convergence and diversity Grid-IGD, set reference points generated by estimating PFs question, based on representative nondominated solutions all grid environment. reduce complexity, Grid-IGD only considers closest solution within neighborhood approximation every point. also possesses other desirable properties, such compliance, immunity dominated/duplicate solutions, no need normalization. experimental studies, verified artificial real obtained five many-objective optimizers. Effects specification behavior are discussed detail theoretically experimentally.