BA (Binary Alphabet)

集成聚类因其在聚类任务中的高性能而在机器学习和数据挖掘中引起了广泛关注。谱聚类是最流行的聚类方法之一，与传统的聚类方法相比具有优越的性能。现有的集成聚类方法通常直接使用基聚类算法的聚类结果进行集成学习，不能很好地利用图拉普拉斯算子在谱聚类中探索到的内在数据结构，从而无法获得理想的聚类结果。在本文中，我们提出了一种基于谱聚类的聚类算法的新集成学习方法。所提出的方法不是直接使用从每个基本谱聚类算法获得的聚类结果，而是通过从每个基本谱聚类算法的谱嵌入中进行集成学习来学习图拉普拉斯算子的鲁棒表示。最后，所提出的方法对从学习图拉普拉斯算子获得的谱嵌入应用 k-means 以获得聚类。合成数据集和真实数据集的实验结果表明，所提出的方法优于其他现有的集成聚类方法。Ensemble clustering has garnered substantial attention in machine learning and data mining due to its outstanding performance in clustering tasks. Spectral clustering is one of the most prevalent clustering methods, which delivers superior performance compared with traditional clustering approaches. Most existing ensemble clustering methods directly utilize the clustering outputs of base clustering algorithms for ensemble learning, yet they fail to sufficiently exploit the intrinsic data structure captured by the graph Laplacian operator in spectral clustering, thus being unable to achieve satisfactory clustering results. In this paper, we propose a novel ensemble learning approach for spectral clustering-based clustering algorithms. The proposed method does not directly employ the clustering results obtained from each base spectral clustering algorithm; instead, it learns a robust representation of the graph Laplacian operator through ensemble learning on the spectral embeddings derived from each base spectral clustering algorithm. Finally, the proposed method applies the k-means algorithm to the spectral embeddings learned from the robust graph Laplacian representation to obtain the final clustering outcomes. Experimental results on both synthetic and real-world datasets demonstrate that the proposed method outperforms other existing state-of-the-art ensemble clustering approaches.

搜集汇总

背景与挑战

背景概述

该数据集用于机器学习和数据挖掘中的聚类任务研究，提出了一种基于谱聚类的集成学习方法，通过谱嵌入学习图拉普拉斯算子的鲁棒表示，并应用k-means进行聚类。实验结果表明，该方法在合成和真实数据集上优于其他现有集成聚类方法。