一种基于T-分布随机近邻嵌入的聚类集成方法

该文将T-分布随机近邻嵌入(TSNE)引入到聚类集成问题中，提出一种基于TSNE的聚类集成方法。首先通过TSNE最小化超图邻接矩阵的行对应的高维数据点与低维映射点分布之间的KL散度，使得高维空间结构在低维空间得以保持，然后在低维空间运行层次聚类算法获得最终的聚类结果。在基准数据集上的实验结果表明: TSNE能够提高层次聚类算法的聚类质量，该文方法获得了优于主流聚类集成方法的结果。

Cluster Ensemble Approach Based on T-distributed Stochastic Neighbor Embedding

T-distributed Stochastic Neighbor Embedding (TSNE) is introduced into cluster ensemble problem and a cluster ensemble approach based on TSNE is proposed. First, TSNE is utilized to minimize Kullback-Leibler divergences between the high-dimensinal points corresponding to the rows of hypergraphs adjacent matrix and the low-dimensional mapping points, which preserves the structure of high-dimensional space in low-dimensional space. Then, a hierarchical clustering algorithm is carried out in the low-dimensional space to obtain the final clustering result. Experimental results on several baseline datasets indicate that TSNE can improve the cluster results of hierarchical clustering algorithm and the proposed cluster ensemble method via TSNE outperforms state-of-the-art methods.