摘要:多视角数据的涌现对传统单视角聚类算法提出了挑战.利用单视角聚类算法独立地对每个视角进行划分，再通过集成机制获取全局划分的方法，人为地割裂了视角之间的内在联系，难以获得理想的聚类效果.针对此问题，提出了一个多视角聚类模型.该模型不仅考虑了视角内的划分质量，还兼顾了视角间的协同学习机制.对于视角内的划分，为了捕捉更为准确的簇内结构信息，采用多代表点的簇结构表示策略；对于视角间的协同学习机制，假设簇中代表点在不同视角下，其代表性保持.因此，在该模型基础上提出了基于代表点一致性约束的多视角模糊聚类算法（multi-view fuzzy clustering with a medoid invariant constraint，简称MFCMddI）.该算法通过最大化两两相邻视角下代表点权重系数的乘积之和来保证代表点一致性.MFCMddI的目标函数可通过引入拉格朗日乘子和KKT条件进行优化.在人工数据集以及真实数据集上的实验结果均表明，该算法相对于所引入的对比算法而言具有一定的优势.



Abstract:As for multi-view datasets, direct integration of partition results of all views obtained by traditional single-view clustering approaches does not improve and even deteriorate the clustering performance since that it does not consider the inner relationship across views. To achieve good clustering performance for multi-view datasets, a multi-view clustering model is proposed, which not only considers the within-view clustering quality but also takes the cross-view collaborative learning into account. With respect to within-view partition, to capture more detailed information of cluster structures, a multi-medoid representative strategy is adopted; as for cross-view collaborative learning, it is assumed that a medoid of a cluster in one view is also a medoid of that cluster in another view. Based on the multi-view clustering model, a multi-view fuzzy clustering approach with a medoid invariant constraint (MFCMddI) is proposed in which the invariantan arbitrary medoid across each pair-wise views is guaranteed by maximizing the product of the corresponding prototype weightsin two views. The objective function of MFCMddI can be optimized by applying the Lagrangian multiplier method and KKT conditions. Extensive experiments on synthetic and real-life datasets show that MFCMddI outperforms the existing state-of-the-art multiview approaches in most cases.



PDF全文下载地址:

http://jos.org.cn/jos/article/pdf/5625