针对不确定数据集进行离群点检测,设计了基于密度的不确定数据的局部离群因子(Uncertain Local Outlier Factor,ULOF)算法。通过建立不确定数据的可能世界模型来确定不确定对象在可能世界中的概率。结合传统的LOF算法推导出ULOF算法,根据ULOF值判断不确定对象的局部离群程度;然后对ULOF算法的效率性和准确性进行了详细分析,提出了基于网格的剪枝策略、k最近邻查询优化来减少数据的候选集;最后通过实验证明了ULOF算法对不确定数据检测的可行性和效率性,优化后的方法有效地提高了异常检测准确率,降低了时间复杂度,改善了不确定数据的异常检测性能。
In order to analyze the capacity stability of the time-varying-propagation and delay-dependent of mobile ad-hoc networks （MANETs）, in this paper, a novel approach is proposed to explore the capacity asymptotic stability for the delay- dependent of MANETs based on non-cooperative game theory, where the delay-dependent conditions are explicitly taken into consideration. This approach is based on the Lyapunov-Krasovskii stability theory for functional differential equations and the linear matrix inequality （LMI） technique. A corresponding Lyapunov-Krasovskii functional is introduced for the stability analysis of this system with use of the descriptor and ＂neutral-type＂ model transformation without producing any additional dynamics. The delay-dependent stability criteria are derived for this system. Conditions are given in terms of linear matrix inequalities, and for the first time referred to neutral systems with the time-varying propagation and delay- dependent stability for capacity analysis of MANETs. The proposed criteria are less conservative since they are based on an equivalent model transformation. Furthermore, we also provide an effective and efficient iterative algorithm to solve the constrained stability control model. Simulation experiments have verified the effectiveness and efficiency of our algorithm.
传统的基于统计的子空间学习算法如主成分分析,通过学习只能得到一系列特征脸,忽略了人脸识别中重要的局部信息(如眼睛、鼻子)。而利用到类别信息的算法如线性判别分析,也会因为小样本问题而有所影响。为了解决这些问题,结合二维偏最小二乘与非负矩阵分解的非负性思想提出二维非负偏最小二乘(Two-Dimensional Nonnegative Partial Least Squares,2DNPLS)算法。其核心思想是在提取人脸特征时加入了非负性约束,使得2DNPLS不仅拥有偏最小二乘算法加入类别信息带来的分类效果,还保留了图像矩阵的内部结构信息,而且还使得到的基矩阵具有非负的局部的可解释性。在ORL,Yale人脸库中的实验结果表明,该算法从时间上和识别率上均优于人脸识别的主流算法。
