基于异常解耦的大规模鲁棒增强集成聚类
Large-scale Robust Enhanced Ensemble Clustering via Outlier Decoupling
论文提出的异常解耦机制（将锚相似矩阵分解为干净成分与残差异常成分，并施加全局互相关惩罚与行级 l2,1 范数）可迁移到视觉任务中，用于处理特征表示中的异常值污染，例如在图像聚类、检索或自监督预训练中提升锚图或相似度矩阵的鲁棒性。
原文依据 RANGE decomposes the anchor similarity matrix into a clean component and a residual outlier component in the anchor space.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论集成聚类旨在从多个基聚类结果中推导出一致的划分。基于锚的方法通过锚构建紧凑的相似性表示，显著提升了计算效率。然而，当异常值污染数据时，重建基聚类结果往往会产生有偏的锚。这些有偏的锚会降低锚相似矩阵的质量，并导致聚类精度下降。为解决该问题，我们提出了一种新方法，称为基于异常解耦的大规模鲁棒增强集成聚类（RANGE）。具体而言，RANGE首先将基聚类结果转换为初始二部图。为增强该二部图的可靠性，RANGE专门针对初始二部图设计了一种高阶模糊增强策略（HFES）。接着，RANGE引入锚矩阵以提升二部图重建过程的计算效率。为提高鲁棒性，RANGE在锚空间中将锚相似矩阵分解为干净成分和残差异常成分。施加全局互相关惩罚以抑制两个成分之间的泄漏，同时对残差部分施加行级 l2,1 范数，将污染限制在少量残差锚方向上。此外，通过对解耦后的异常结构应用异常检测器，RANGE可扩展至异常检测任务。因此，RANGE形成了一个跨任务的通用框架。大量实验表明，RANGE在聚类有效性和异常检测方面均表现出优越性能。
英文原文摘要 1648 字符
Ensemble clustering aims to derive a consensus partition from multiple base clustering results. Anchor-based methods construct compact similarity representations via anchors, substantially improving computational efficiency. However, when outliers contaminate the data, reconstructing the base clustering results often yields biased anchors. These biased anchors degrade the quality of the anchor similarity matrix and lead to a decline in clustering accuracy. To address this issue, we propose a novel method called large-scale robust enhanced ensemble clustering via outlier decoupling (RANGE). Specifically, RANGE first converts the base clustering results into an initial bipartite graph. To enhance the reliability of this bipartite graph, RANGE designs a high-order fuzzy enhancement strategy (HFES) specifically for initial bipartite graphs. Next, RANGE introduces an anchor matrix to improve the computational efficiency of the bipartite graph reconstruction process. To improve robustness, RANGE decomposes the anchor similarity matrix into a clean component and a residual outlier component in the anchor space. A global cross-correlation penalty is imposed to suppress leakage between the two components, while a row-wise l_ 2,1 -norm on the residual part confines contamination to a small number of residual anchor directions. Moreover, by applying outlier detectors to the decoupled outlier structure, RANGE can be extended to the outlier detection task. Consequently, RANGE forms a cross-task general framework. Extensive experiments indicate that RANGE delivers superior performance in both clustering validity and outlier detection.