Multivariate Optimization Algorithm Implemented In C ++ And Multi-modal Optimization Study

In this paper, a novel multi-group variant optimization algorithm by tracking the historical information—multi-variant optimization algorithm(MOA) is introduced by analyzing the constraints when the traditional swarm intelligence algorithm was designed and combining the advantage of compute and storage. The purpose of the new algorithm is to solve the optimization problem from a new perspective by making full use of the characteristic of modern computer memory with high capacity and read speed. The main idea of MOA is to search the solution space comprehensively through global and local search alternately. In order to store and share the information gained during the search process effectively, a special structure table is constructed and the search atom is positioned at the right position by following the operating rules of the table.Firstly, the construction and operation rules of structure table in MOA is described in detail and the flowchart of the algorithm was given. Then some properties of MOA are discussed briefly. After analysis how the support of low level memory allocation and pointer arithmetic makes C++the best choice to implement the algorithm, we construct a new MOA platform in C++. In the part of implementation, we describes abstract type and interface of MOA in the form of data schema and makes the structure table works as the algorithm expected by define the functionality through the interface. The new C++platform was integrated into the Matlab through mex mixed programming, this expanded the field of MOA usage. Finally, the multi-group property of MOA and how the property made the algorithm a perfect candidate for multi-modal optimization without the need of any extra mechanism is discussed. Then the paper analyzed the difference between MOA and mainstream multi-modal algorithm. In the part of experimentation, seven classic multi-modal benchmark functions with different property are introduced for comprehensive analysis of the algorithm in multi-modal optimization. The feasibility of MOA was validated by observing the position of search atoms in the solution space of each function. The results also shows that MOA can still keep high stability in different environment. At last, MOA and three niche PSO algorithms which were designed for multimodal optimization were applied to seven functions under the same condition for statistical analysis. The performance of four algorithms was analyzed from three performance indicators:accuracy, success rate and convergence speed. The results show that MOA is an effective multi-modal optimization.