期刊文献+
共找到4篇文章
< 1 >
每页显示 20 50 100
结合密度峰值和改进自然邻居的自训练算法 被引量:2
1
作者 吕佳 刘强 李帅军 《南京大学学报(自然科学版)》 CAS CSCD 北大核心 2022年第5期805-815,共11页
自训练算法是一种经典的半监督学习算法.针对自训练算法在训练时很难有效地发现高置信度无标记样本、训练过程中样本容易被错误标记等问题,提出了结合密度峰值和改进自然邻居的自训练算法.该算法利用密度峰值聚类算法的密度信息和由自... 自训练算法是一种经典的半监督学习算法.针对自训练算法在训练时很难有效地发现高置信度无标记样本、训练过程中样本容易被错误标记等问题,提出了结合密度峰值和改进自然邻居的自训练算法.该算法利用密度峰值聚类算法的密度信息和由自然邻居算法得到的自然邻居集来计算样本的离群指数,结合表示全局特征的全局密度和表示局部特征的离群指数设定样本的权重并剔除离群点.在自训练算法逐步扩充有标记样本集的过程中,利用加权的自然邻居噪声过滤器找出误标记样本,并通过计算样本的不确定性找出高不确定性的样本,由主动学习给出准确标记.提出的算法不仅降低了离群点对算法性能的影响,而且有效解决了错误标记的问题,处理了不易标记的边界样本.实验表明,该算法在UCI基准数据集和噪声数据集上均取得了理想的效果. 展开更多
关键词 半监督学习 自训练算法 密度峰值聚类 自然邻居 噪声过滤器
在线阅读 下载PDF
多模型信息融合Wiener状态估值器
2
作者 杨立新 冉陈键 邓自立 《科学技术与工程》 2007年第24期6278-6284,6294,共8页
对多模型多传感器线性离散定常随机系统,应用现代时间序列分析方法,基于自回归滑动平均(ARMA)新息模型和白噪声估计理论,根据按矩阵加权、按对角阵加权和按标量加权三种最优融合规则,提出了系统公共状态的三种最优加权融合Wiener估值器... 对多模型多传感器线性离散定常随机系统,应用现代时间序列分析方法,基于自回归滑动平均(ARMA)新息模型和白噪声估计理论,根据按矩阵加权、按对角阵加权和按标量加权三种最优融合规则,提出了系统公共状态的三种最优加权融合Wiener估值器。它们的精度高于每个局部估值器的精度,且可统一处理融合滤波、预报和平滑问题。为计算最优加权,提出计算局部估计误差互协方差公式。它们可用于带ARMA有色观测噪声系统状态融合滤波问题。一个跟踪系统MonteCarlo仿真例子说明其有效性。 展开更多
关键词 多传感器信息融合 不同局部动态模型 加权状态融合 WIENER滤波器 噪声过滤器 现代时间序列分析方法
在线阅读 下载PDF
Designing and implementation of variance-dependent Goldstein radar interferogram filtering 被引量:1
3
作者 王琪洁 张小虎 +1 位作者 朱建军 胡俊 《Journal of Central South University》 SCIE EI CAS 2014年第8期3295-3301,共7页
The variance-dependent Goldstein radar interferogram filter takes into account the information of both interferometric coherence and multilook factors,and can produce very consistent results for interferograms generat... The variance-dependent Goldstein radar interferogram filter takes into account the information of both interferometric coherence and multilook factors,and can produce very consistent results for interferograms generated under a wide variety of multilook factors and with very different noise level.However,the filter is a bit complicated and its application is still very limited.We present the designing and implementation of the variance-dependent Goldstein radar interferogram filtering,emphasizing on the logic flow,the generation of look-up table,the determination of filtering parameter,and the handling of edge information loss.Experiments with real interferograms are provided to demonstrate the applications of the designed filtering.Comparisons with the result of the coherence-dependent Goldstein filter show that improvements from 18.4% to 36.9% are achieved when the variance-dependent filter is used,and the noisier the interferogram,the greater the improvement. 展开更多
关键词 synthetic aperture radar interferometry (InSAR) Goldstein filter variance-dependent radar COHERENCE multilook factor
在线阅读 下载PDF
A novel reaching law approach of quasi-sliding-mode control for uncertain discrete-time systems 被引量:4
4
作者 袁雷 沈建清 +1 位作者 肖飞 王恒利 《Journal of Central South University》 SCIE EI CAS 2012年第9期2514-2519,共6页
A novel discrete-time reaching law was proposed for uncertain discrete-time system,which contained process noise and measurement noise.The proposed method reserves all the advantages of discrete-time reaching law,whic... A novel discrete-time reaching law was proposed for uncertain discrete-time system,which contained process noise and measurement noise.The proposed method reserves all the advantages of discrete-time reaching law,which not only decreases the band width of sliding mode and strengthens the system robustness,but also improves the dynamic performance and stability capability of the system.Moreover,a discrete-time sliding mode control strategy based on Kalman filter method was designed,and Kalman filter was employed to eliminate the influence of system noise.Simulation results show that there is no chattering phenomenon in the output of controller and the state variables of controlled system,and the proposed algorithm is also feasible and has strong robustness to external disturbances. 展开更多
关键词 discrete-time system quasi-sliding-mode control discrete-time reaching law Kalman filter
在线阅读 下载PDF
上一页 1 下一页 到第
使用帮助 返回顶部