An Ameliorated Harmony Search Algorithm With Hybrid Convergence Mechanism

The metaheuristic optimization algorithm Harmony search (HS), which imitates the process of music improvisation, is becoming widely used due to its simplicity and easy operation. However, the basic HS has shortcomings of low optimization accuracy and risk of easy falling into local optimum. To overcome these problems, this article develops an ameliorated harmony search algorithm with hybrid convergence mechanism, namely AHS-HCM. For the new method, the so-called hybrid convergence mechanism mainly includes two important schemes. The first scheme is to introduce a convergence coefficient in the harmony improvisation to further adjust the optimization performance, which can improve the final accuracy. The second scheme is to put forward a non-linear convergence domain for the global exploration, it also helps the optimization accuracy and efficiency. Besides, the operation process of new harmony variables generation is modified to enrich the search behavior and contribute to find the global optimum. Fifteen typical CEC's benchmark functions are selected for experiments, and the results clearly proved the effectiveness of the AHS-HCM. It is shown that, in most cases, the proposed AHS-HCM algorithm is superior to other HS variants as well as some similar famous population-based algorithms in terms of optimization accuracy and stability.

Automated summary

The improved harmony search algorithm AHS-HCM overcomes the shortcomings of basic HS in terms of low optimization accuracy and risk of falling into local optimum by introducing a hybrid convergence mechanism, and experiments have shown its effectiveness compared to other HS variants and population-based algorithms.