Improving the Adaptive Properties of LSHADE Algorithm for Global Optimization

Differential Evolution is a really challenging algorithm in the field of computational intelligence and has proved its worth in solving various real-world optimization problems. The algorithm since its inception has been enhanced to improve its competitiveness and various new versions have been designed. In present work, the properties of an enhanced version of DE namely LSHADE algorithm are enhanced and new version namely SALSHADE is proposed. The newly proposed version consists of three major modifications that is, i) exponentially decreasing crossover rate, ii) linearly decreasing scaling factor and iii) frequency component is enhanced by using L/evy distributed step size. These three modifications have been added and experimental analysis is done on CEC2017 benchmark problems to prove its worth. The new proposed SALSHADE algorithm is compared with SaDE, JADE, SHADE, MVMO, CVsin, CV1.0, LSHADEcnEpSin and other algorithms. Further, experimental results show that SALSHADE is highly competitive and is a potential candidate for becoming state-of-the-art.