Unbalanced budget distribution for automatic algorithm configuration

Optimization algorithms often have several critical setting parameters and the improvement of empirical performance these depends on tuning them. Manually configuration such is a tedious task that results in unsatisfactory outputs. Therefore, automatic algorithm frameworks been proposed to regulate given for series problem instances. Although developed perform very well deal with various problems, however, there still trade-off between accuracy budget requirements need be addressed. This work investigates unbalanced distribution different configurations problem. Inspired by bandit-based approaches, main goal find better substantially improves target while using smaller run time budget. In this work, non-dominated sorting genetic II employed as jMetalPy software platform multimodal multi-objective optimization (MMO) test suite CEC’2020 used set problems. We did comprehensive comparison other known methods including random search, Bayesian optimization, sequential model-based (SMAC), iterated local search parameter space (ParamILS), racing (irace), many-objective (MAC) methods. order characterize, validate evaluate methods, hypervolume (HV), generational distance, epsilon indicator ( $$I_{{\epsilon }^{+}}$$ ) are indicators. The experimental interestingly proved efficiency approach minimum competitors.