聚类有效性评价综述

在聚类分析应用中,迫切需要一种客观公正的质量评价方法来评判聚类结果的有效性。为此,从外部评价法、内部评价法和相对评价法三个方面,归纳综述了常用的聚类有效性评价方法,并讨论了模糊聚类评价法和聚类最佳类别数的自动确定问题。

聚类组合研究的新进展

作为目前聚类分析的新兴研究热点,聚类组合方法能将两种或多种聚类方法集成起来以改善其性能。从聚类多样性和共识函数两方面综述了最新研究进展,探讨将神经网络组合的思想用于聚类组合。最后指出了将来可能的研究方向。

基于自适应谐振理论的Web文档聚类集成方法

提出了基于自适应谐振理论的Web文档聚类集成方法.该方法先用蚁群算法对Web文档进行聚类,再用ART神经网络对聚类结果集成.实验结果表明,集成后的聚类综合质量高于集成前的聚类综合质量.

Unknown Inputs Observer for a Class of Nonlinear Uncertain Systems: An LMI Approach

This paper deals with the simultaneous estimation of states and unknown inputs for a class of Lipschitz nonlinear systems using only the measured outputs. The system is assumed to have bounded uncertainties that appear on both the state and output matrices. The observer design problem is formulated as a set of linear constraints which can be easily solved using linear matrix inequalities (LMI) technique. An application based on manipulator arm actuated by a direct current (DC) motor is presented to evaluate the performance of the proposed observer. The observer is applied to estimate both state and faults.