An approach for initializing the random adaptive grouping algorithm for solving large-scale global optimization problems

Abstract

Many real-world optimization problems deal with high dimensionality and are known as large-scale global optimization (LSGO) problems. LSGO problems, which have many optima and are not separable, can be very challenging for many heuristic search algorithms. In this study, we have proposed a novel two-stage hybrid heuristic algorithm, which incorporates the coordinate descent algorithm with the golden-section search (CDGSS) and the random adaptive grouping for cooperative coevolution of the Self-adaptive Differential Evolution with Neighborhood Search (DECC-RAG) algorithm. At the first stage, the proposed algorithm roughly scans the search space for a better initial population for the DECC-RAG algorithm. At the second stage, the algorithm uses the DECC-RAG framework for solving the given LSGO problem. We have evaluated the proposed approach (DECC-RAG1.1) with 15 most difficult LSGO problems from the IEEE CEC’2013 benchmark set. The experimental results show that DECC-RAG1.1 outperforms the standard DECC-RAG and some the state-of-the-art LSGO algorithms.

A Novel Memetic Algorithm Based on Multiparent Evolution and Adaptive Local Search for Large-Scale Global Optimization

“…The hybridization between global and local search algorithms has been experimentally proven to provide better search performance [ 33 ]. Many examples have proved the effectiveness of this strategy, including many difficult problems, such as multimodal optimization [ 33 , 34 ], large-scale global optimization [ 5 – 19 ], combinatorial optimization [ 32 , 35 ], single-objective optimization [ 36 , 37 ], and multiobjective optimization [ 38 , 39 ]. For large-scale global optimization problems, most of the best-performing algorithms were hybrid algorithms combining global and local search (e.g., SHADE-ILS, MLSHADE-SPA, and MOS).…”

“…CBCC-RDG3 [ 17 ] is also a modified version of the CC algorithm that modifies the recursive differential grouping method to reduce the overlapping problems. TPHA [ 18 ] and DECC-RAG1.1 [ 19 ] are two-phase hybrid algorithms that use the CC framework.…”