A Budget-Controlled In Vitro Fertilization Memetic Hybrid for SPEA2 on Bi- and Tri-Objective Problems

Abstract The Strength Pareto Evolutionary Algorithm 2 (SPEA2) remains a standard baseline in multiobjective optimization, but its density-based truncation can remove offspring generated by local intensification. We study a budget-controlled memetic hybrid that inserts an in vitro fertilization phase into SPEA2 while preserving the host's generation-level evaluation budget. The implementation analyzed here combines dissimilar-father selection with a collective cycle-continuation rule so that intensification is retained only when it improves the population under SPEA2's archive dynamics. A three-phase tuning pipeline on 12 representative problem-objective configurations defines the default setting. The resulting method is evaluated on 39 out-of-sample synthetic instances, the full 51-instance synthetic suite, and three constrained engineering problems. Against canonical SPEA2 over 60 runs per configuration, the method achieves Holm--Bonferroni-corrected inverted generational distance win/loss/tie counts of 19/2/3 on out-of-sample bi-objective instances and 12/0/3 on out-of-sample tri-objective instances; on the full synthetic suite the corresponding counts are 22/3/3 and 15/1/7. Hypervolume provides secondary support. Improvements concentrate on regular Pareto-front geometries, whereas disconnected or irregular fronts can weaken or reverse the effect because intensification-generated density interacts unfavorably with SPEA2 truncation. The results therefore support a conditional, host-specific memetic improvement over SPEA2 at moderate objective counts rather than a universal replacement.