摘要
提出了一种负熵最小化加权最小二乘支持向量机分类模型,并应用于水下底质识别任务.该模型在原始最小二乘支持向量机(Least Square Support Vector Machine,LSSVM)基础上引入权重,通过权重分布的负熵最小化调节和控制权重的稀疏度,然后使用该稀疏分布权重进一步进行加权LSSVM再学习,从而实现对原始LSSVM分类边界的调整优化.将负熵最小化加权LSSVM应用于水下钴结壳底质识别,实验结果表明,该负熵最小化加权LSSVM能显著减小钴结壳错判率和识别正确率,有效提高底质识别效果.
A weighted least squares support vector machine based on negative entropy minimization is proposed, which is used on the task of underwater bottom material recognition. The proposed classifier model is learned on the results of original Least Squares Support Vector Machine（LSSVM）. Through negative entropy minimization of distribution of weights, it controls and adjusts sparseness of sampling weights effectively. And then, by using these sparse weights, a weighted LSSVM is retrained to adjust the original classification border of LSSVM. Finally, the proposed model is used on underwater cobalt-rich crusts recognition. Experiment results show that using the proposed weighted LSSVM based on negative entropy minimization can decrease the misclassification rate and recognition correct rates significantly and improve the effectiveness of bottom material recognition.
作者
杨勃
邵泉铭
YANG Bo SHAO Quanming(College of Information and Communication Engineering, Hunan Institute of Science and Technology, Yueyang 414006, China)
出处
《湖南理工学院学报（自然科学版）》
CAS
2017年第3期27-32,66,共7页
Journal of Hunan Institute of Science and Technology(Natural Sciences)
基金
湖南省教育厅科学研究重点资助项目(17A089)
关键词
加权最小二乘支持向量机
负熵
稀疏权重
钴结壳识别
底质识别
weighted LSSVM, negative entropy, sparse weights, cobalt-rich crusts recognition, bottom material recognition
