期刊文献+
共找到3篇文章
< 1 >
每页显示 20 50 100
局部密度最小不确定性的SVM样本选择算法
1
作者 周玉 刘虹瑜 +2 位作者 李京京 丁红强 白磊 《哈尔滨工业大学学报》 北大核心 2025年第8期45-56,共12页
为解决支持向量机(SVM)在分类时通常含有大量的冗余样本,从而导致面对较大规模数据集时SVM计算复杂度受到限制的问题,提出一种局部密度最小不确定性的SVM样本选择算法。该方法对决策面影响较大的边界数据进行有效选择,通过提取可能含有... 为解决支持向量机(SVM)在分类时通常含有大量的冗余样本,从而导致面对较大规模数据集时SVM计算复杂度受到限制的问题,提出一种局部密度最小不确定性的SVM样本选择算法。该方法对决策面影响较大的边界数据进行有效选择,通过提取可能含有支持向量的训练样本,降低计算开销,进而提高SVM性能。首先,计算训练样本的K互近邻个数与高斯核密度估计。其次,将K互近邻个数与高斯核密度估计进行加和得到每个样本点的K局部密度并获取密度矩阵。然后,利用局部密度不确定性平衡优化方法,将密度矩阵进行三值映射后使不确定性改变量达到最小时得到最优阈值,并划分密度矩阵为中心数据与边界数据。最后,提取边界数据并作为SVM的训练样本建立分类模型。结果表明:利用该方法在UCI数据集上与其他6种常用样本选择方法进行实验对比,以准确率、保存率作为性能指标,文中提出的算法可以迅速划分中心数据与边界数据并删除大量冗余的训练样本,有效降低SVM的训练负担的同时提高了分类性能。 展开更多
关键词 支持向量机(SVM) 样本选择 局部密度 不确定性平衡 分类
在线阅读 下载PDF
Design for robust stabilization of nonlinear systems with uncertain parameters 被引量:1
2
作者 赖旭芝 文静 吴敏 《Journal of Central South University of Technology》 EI 2004年第1期102-104,共3页
Based on Lyapunov stability theory, a design method for the robust stabilization problem of a class of nonlinear systems with uncertain parameters is presented. The design procedure is divided into two steps: the firs... Based on Lyapunov stability theory, a design method for the robust stabilization problem of a class of nonlinear systems with uncertain parameters is presented. The design procedure is divided into two steps: the first is to design controllers for the nominal system and make the system asymptotically stabi1ize at the expected equilibrium point; the second is to construct closed-loop nominal system based on the first step, then design robust controller to make the error of state between the origina1 system and the nominal system converge to zero, thereby a dynamic controller with the constructed closed-loop nominal system served as interior dynamic is obtained. A numerical simulation verifies the correctness of the design method. 展开更多
关键词 nonlinear system robust stabilization Lyapunov stability
在线阅读 下载PDF
Similarity measure design on overlapped and non-overlapped data
3
作者 LEE Sang-hyuk SHIN Seung-soo 《Journal of Central South University》 SCIE EI CAS 2013年第9期2440-2446,共7页
Similarity measure design on non-overlapped data was carried out and compared with the case of overlapped data.Unconsistant feature of similarity on overlapped data to non-overlapped data was provided by example.By th... Similarity measure design on non-overlapped data was carried out and compared with the case of overlapped data.Unconsistant feature of similarity on overlapped data to non-overlapped data was provided by example.By the artificial data illustration,it was proved that the conventional similarity measure was not proper to calculate the similarity measure of the non-overlapped case.To overcome the unbalance problem,similarity measure on non-overlapped data was obtained by considering neighbor information.Hence,different approaches to design similarity measure were proposed and proved by consideration of neighbor information.With the example of artificial data,similarity measure calculation was carried out.Similarity measure extension to intuitionistic fuzzy sets(IFSs)containing uncertainty named hesitance was also followed. 展开更多
关键词 similarity measure overlapped data non-overlapped data intuitionistic data
在线阅读 下载PDF
上一页 1 下一页 到第
使用帮助 返回顶部