aiqbench analysis
This article addresses a critical vulnerability in Graph Neural Networks where contaminated training data degrades unsupervised anomaly detection performance. The proposed DeNoise framework utilizes adversarial objectives and contrastive learning to generate robust embeddings that resist noise interference, though the authors have withdrawn the paper due to identified issues in their original experimental evaluation. For practitioners relying on graph-structured data integrity, this highlights the ongoing necessity for rigorous validation of training set purity before deployment.
DeNoise: Learning Robust Graph Representations for Unsupervised Graph-Level Anomaly Detection
arXiv:2511.04086v2 Announce Type: replace-cross Abstract: With the rapid growth of graph-structured data in critical domains, unsupervised graph-level anomaly detection (UGAD) has become a pivotal task. UGAD seeks to identify entire graphs that deviate from normal behavioral patterns. However, most Graph Neural Network (GNN) approaches implicitly assume that the training set is clean, containing only normal graphs, which is rarely true in practice. Even modest contamination by anomalous graphs can distort learned representations and sharply degrade performance. To address this challenge, we propose DeNoise, a robust UGAD framework explicitly designed for contaminated training data. It jointly optimizes a graph-level encoder, an attribute decoder, and a structure decoder via an adversarial objective to learn noise-resistant embeddings. Further, DeNoise introduces an encoder anchor-alignment denoising mechanism that fuses high-information node embeddings from normal graphs into all graph embeddings, improving representation quality while suppressing anomaly interference. A contrastive learning component then compacts normal graph embeddings and repels anomalous ones in the latent space. Extensive experiments on eight real-world datasets demonstrate that DeNoise consistently learns reliable graph-level representations under varying noise intensities and significantly outperforms state-of-the-art UGAD baselines.
Abstract:With the rapid growth of graph-structured data in critical domains, unsupervised graph-level anomaly detection (UGAD) has become a pivotal task. UGAD seeks to identify entire graphs that deviate from normal behavioral patterns. However, most Graph Neural Network (GNN) approaches implicitly assume that the training set is clean, containing only normal graphs, which is rarely true in practice. Even modest contamination by anomalous graphs can distort learned representations and sharply degrade performance. To address this challenge, we propose DeNoise, a robust UGAD framework explicitly designed for contaminated training data. It jointly optimizes a graph-level encoder, an attribute decoder, and a structure decoder via an adversarial objective to learn noise-resistant embeddings. Further, DeNoise introduces an encoder anchor-alignment denoising mechanism that fuses high-information node embeddings from normal graphs into all graph embeddings, improving representation quality while suppressing anomaly interference. A contrastive learning component then compacts normal graph embeddings and repels anomalous ones in the latent space. Extensive experiments on eight real-world datasets demonstrate that DeNoise consistently learns reliable graph-level representations under varying noise intensities and significantly outperforms state-of-the-art UGAD baselines.
Withdrawn by the authors because issues were identified in the original data and experimental evaluation, requiring substantial revisions that affect the reported results and conclusions.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2511.04086 [cs.LG]
(or arXiv:2511.04086v2 [cs.LG] for this version)
Full text under the publisher's open licence · 216 words
Related coverage
PIRAMID: Progress and Plans
LessWrong · Jul 27, 2026
Robust and Efficient AI Frameworks for Scalable Material Design and Property Prediction
arXiv · Sep 17, 2026
D-CLOT: Double Closed Loop Optimal Transport for Unsupervised Action Segmentation
arXiv · Aug 7, 2026
Unsupervised Anatomical Feature Learning via Diffusion Models: Enhanced Medical Image Segmentation with Denoising Diffusion Probabilistic Models
arXiv · Aug 27, 2026