Research map / Graph learning / Graph self-supervised learning
Towards Multiple Missing Values-resistant Unsupervised Graph Anomaly Detection
Jiazhen Chen, Xiuqin Liang, Sichao Fu, Zheng Ma, Weihua Ou
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add structure reconstruction pathway (PPR diffusion) (ablation) | M2V-UGAD · Cora, Citeseer, Books, Disney (30% missing) | AUROC: about 10-30% AUROC lower without it | Improved |
| add feature imputation pathway (ablation) | M2V-UGAD · Cora, Citeseer, Books, Disney (30% missing) | AUROC: about 3-10% AUROC lower without it | Improved |
| add pseudo-anomaly generation (ablation) | M2V-UGAD · Cora, Citeseer, Books, Disney (30% missing) | AUROC: about 10-35% AUROC lower without it | Improved |
| add feature imputation loss Lfeat (ablation) | M2V-UGAD · Cora, Citeseer, Books, Disney (30% missing) | AUROC: about 2-4% AUROC lower without it | Improved |
| add reconstruction loss Lrecon (ablation) | M2V-UGAD · Cora, Citeseer, Books, Disney (30% missing) | AUROC: about 2-4% AUROC lower without it | Improved |
| M2V-UGAD vs ADA-GAD with ASD-VAE imputation | M2V-UGAD · Books (30% missing) | AUROC: 0.70 -> 0.63 | Hurt |
| M2V-UGAD vs best impute-then-detect baseline | M2V-UGAD · Cora (30% missing) | AUROC: 0.84 -> 0.93 | Improved |
| M2V-UGAD vs best impute-then-detect baseline | M2V-UGAD · Flickr (30% missing) | AUROC: 0.74 -> 0.93 | Improved |
| raise missing rate from 10% to 50% | M2V-UGAD · Cora | AUROC: 0.93 -> 0.92 | No clear effect |
| raise missing rate from 10% to 50% for ADA-GAD | ADA-GAD · Cora | AUROC: 0.86 -> 0.83 | Hurt |
| increase imputation loss weight alpha above 0.01 (ablation) | M2V-UGAD · Cora, Citeseer, Books, Disney | AUROC | Hurt |
| increase pseudo-anomaly ratio eta from 0.1 to 0.5 (ablation) | M2V-UGAD · Disney | AUROC | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ADA-GAD — ADA-GAD: Anomaly-Denoised Autoencoders for Graph Anomaly Detection (AAAI 2024) · better on 18 of 19 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- UniGAD: Unifying Multi-level Graph Anomaly Detection (NeurIPS 2024 · similarity 0.92)
- ARC: A Generalist Graph Anomaly Detector with In-Context Learning (NeurIPS 2024 · similarity 0.91)
- Open-Set Graph Anomaly Detection via Normal Structure Regularisation (ICLR 2025 · similarity 0.91)
- Generative Semi-supervised Graph Anomaly Detection (NeurIPS 2024 · similarity 0.90)
- Unifying Unsupervised Graph-Level Anomaly Detection and Out-of-Distribution Detection: A Benchmark (ICLR 2025 · similarity 0.90)
- DiffGAD: A Diffusion-based Unsupervised Graph Anomaly Detector (ICLR 2025 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
