Testing MVMO on learning-based real-parameter single objective benchmark optimization problems

Mean-variance mapping optimization (MVMO) is an emerging evolutionary algorithm, which adopts a single-solution based approach and performs evolutionary operations within a normalized range of the search for all optimization variables. MVMO uses a special mapping function for mutation operation, which allows a controlled shift from exploration priority at early stages of the search process to exploitation at later stages. Recently, the MVMO has been extended to a population-based and hybrid variant denoted as MVMO-SH, which includes strategies for local search and multi-parent crossover. This paper provides an study on the performance of MVMO-SH on the IEEE-CEC 2015 competition test suite on learning-based real-parameter single objective optimization. Experimental results evidence the effectiveness of MVMO-SH for successfully solving different optimization problems with different mathematical properties and dimensionality.