Adversarial-Robustness-Guided Graph Pruning
Yongyu Wang
TL;DR
The paper tackles the challenge of learning robust, scalable graph topologies for spectral methods by introducing a four-phase adversarial-robustness-guided pruning framework. It combines an initial k-NN construction, spectral embedding, and a Spade-based robustness evaluation to identify and remove non-robust edges, yielding a pruned graph optimized for spectral clustering. Empirical results on benchmark datasets demonstrate improved clustering accuracy and computational efficiency over traditional Laplacian-based graph learning approaches. This approach enables robust graph topology learning suitable for large-scale graph-based tasks in data mining and beyond.
Abstract
Graph learning plays a central role in many data mining and machine learning tasks, such as manifold learning, data representation and analysis, dimensionality reduction, clustering, and visualization. In this work, we propose a highly scalable, adversarial-robustness-guided graph pruning framework for learning graph topologies from data. By performing a spectral adversarial robustness evaluation, our method aims to learn sparse, undirected graphs that help the underlying algorithms resist noise and adversarial perturbations. In particular, we explicitly identify and prune edges that are most vulnerable to adversarial attacks. We use spectral clustering, one of the most representative graph-based machine learning algorithms, to evaluate the proposed framework. Compared with prior state-of-the-art graph learning approaches, the proposed method is more scalable and significantly improves both the computational efficiency and the solution quality of spectral clustering.