Hierarchical Anchor Graph Learning for Multi-View Clustering

Multi-view clustering (MVC) is a fundamental task in heterogeneous data analysis, where anchor-based graph methods are widely adopted for their computational efficiency. However, existing approaches typically utilize static, single-layer anchors, failing to capture the multi-granularity nature of complex data. Drawing inspiration from hierarchical human cognition, we propose a hierarchical anchor graph learning method, termed HAG-MVC, a novel framework that organizes multi-view data as a multi-level pyramid. Unlike conventional one-shot anchor generation methods, HAG-MVC introduces a multi-level co-evolution mechanism, where anchors and graph structures are iteratively refined together to capture semantics from fine-to-coarse granularities. Moreover, HAG-MVC offers a transparent abstraction architecture as an alternative to black-box deep clustering: by maintaining all anchors within the original feature space, it enables explicit inspection of the abstraction process, ensuring inherent interpretability. Extensive experiments on benchmark datasets demonstrate that HAG-MVC consistently outperforms state-of-the-art methods. Beyond MVC, this work provides a scalable and trustworthy paradigm for hierarchical knowledge representation in broad machine learning tasks.

Many real-world datasets describe the same objects in several different ways. For example, a video may include visual, audio, and text information, while a medical record may include different types of test results. A common goal is to automatically group similar objects without relying on human labels. Existing efficient methods often choose a fixed set of representative examples before learning begins, which can miss how simple local patterns combine into higher-level meanings. We propose HAG-MVC, a method that organizes data in layers, similar to how people first notice local details and then form broader concepts. Instead of relying on fixed representatives, our method updates these representatives and their connections together, allowing the grouping process to move from fine details to more abstract patterns. Because the representatives stay in the original data space, researchers can inspect how groups are formed rather than treating the system as a black box. Experiments on a range of datasets show that this layered and transparent approach improves grouping accuracy while remaining efficient for large datasets.