文章摘要
杨宏晖,王芸,戴健.用于水声目标分类的加权免疫克隆样本选择算法[J].声学技术,2013,32(1):46~49
用于水声目标分类的加权免疫克隆样本选择算法
A new weighted immune clonal algorithm for underwater acoustic target classification
投稿时间:2012-05-28  修订日期:2012-07-19
DOI:10.3969/j.issn1000-3630.2013.01.011
中文关键词: 加权免疫克隆样本选择  水声目标分类  样本选择
英文关键词: weight immune clonal instance selection  underwater acoustic target classification  instance selection
基金项目:西北工业大学本科毕业设计重点扶持项目(20120305)
作者单位E-mail
杨宏晖 西北工业大学航海学院,西安 710072  
王芸 西北工业大学航海学院,西安 710072  
戴健 西北工业大学航海学院,西安 710072 daij060766@mail.nwpu.edu.cn 
摘要点击次数: 1057
全文下载次数: 1263
中文摘要:
      提出了一种新的用于水声目标分类的加权免疫克隆样本选择算法(weight Immune Clonal Instance Selection,wICISA)。算法利用Adaboost算法给予每个样本一权值,每代中根据样本权值计算抗体亲合度和克隆数,且根据克隆复制、抗体更新(海明距离交叉和加权简化最近邻变异)和克隆选择操作指导种群进化。提取了实测3类水声目标的时域波形结构特征、小波分析特征和听觉谱特征,进行样本选择和分类仿真实验,结果表明:wICISA可以选出有效样本子集,使样本数目减少82%左右,并且支持向量机分类器的正确分类率能提高约2%;wICISA具有较好的收敛性、稳定性,所得优化样本子集具有较好的泛化能力且能明显减少分类的时间。
英文摘要:
      A new adaptive weighted immune clonal instance selection algorithm (wICISA) for underwater acoustic targets classification is proposed in this paper. The wICISA gets the weight of instances through the Adaboosting algorithm, in each iteration, the affinity of antibodies and the number of clones of antibodies are calculated with the weight of instance. And then, new generations are generated through repetitive application of clonal copy, antibody update (Hamming distance crossover and Weight Reduced Nearest Neighbor mutation) and clonal selection operation. The time wave structure features, wavelet analysis features and auditory spectrum features are extracted from 3 classes of underwater targets, and are used in instance selection and classification experiments. Experimental results show that wICISA can select the subsets of efficient instances with about 2% increase in the accuracy of SVM classifier when the number of instances decreases by about 82%. wICISA has good convergence and stability, and the instance subset obtained by wICISA achieves with good generalizability, and it can reduce the classification time remarkably.
查看全文   查看/发表评论  下载PDF阅读器
关闭