期刊
EXPERT SYSTEMS WITH APPLICATIONS
卷 229, 期 -, 页码 -出版社
PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.eswa.2023.120533
关键词
Artificial bee colony algorithm; Population distribution information; Cumulative covariance matrix; Adaptive selection mechanism; Hearing loss detection
This paper proposes a new ABC algorithm, called CCoM-ABC, which improves the performance of non-separable problems by incorporating population distribution information from previous generations in the evolutionary process. The algorithm applies a cumulative covariance matrix mechanism and an adaptive selection mechanism to enhance the search ability. Experimental results demonstrate that CCoM-ABC outperforms state-of-the-art ABC variants and EAs in terms of convergence speed, solution accuracy, and numerical scalability.
Artificial bee colony (ABC) algorithms which are applied to various complex problems for their competitive performance and simple structures are a relatively popular paradigm of evolutionary algorithms (EAs). Although many variants based on the basic ABC algorithm have been developed, most variants have limited performance on non-separable problems. This is mainly because these variants usually focus on only individual quality information, but neglect population distribution information. To address this limitation, in this paper, we propose a new ABC with a cumulative covariance matrix mechanism (CCoM-ABC), in which the population distribution information in all previous generations is involved to guide the evolution process. Specifically, the CCoM-ABC algorithm applies a cumulative covariance matrix (CCoM) mechanism by which the population information of every generation is involved to improve the probability of finding the optimum. It also applies an adaptive selection mechanism by which the proper coordinate system used to generate candidate solutions can be dynamically selected between a natural one and an eigen one during the search process. Furthermore, for improving the search ability, a new search equation is adopted in the scout bees phase. A series of experiments are conducted on the CEC2014 to evaluate the performance of the proposed CCoM-ABC algorithm, and the results show that the CCoM-ABC outperforms state-of-the-art ABC variants and EAs in terms of convergence speed, solution accuracy, and numerical scalability. For validating the practicality of the proposed CCoM-ABC algorithm, we further apply it to optimize the parameters of the neural network for hearing loss (HL) detection. This study involves 60 brain images consisting of 20 left-sided HL, 20 right-side HL, and 20 healthy control images. The results show that this model outperforms state-of-the-art related methods in terms of overall accuracy by about 2%.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据