# Rethinking Reconstruction-based Graph-Level Anomaly Detection: Limitations and a Simple Remedy

arXiv: https://arxiv.org/abs/2410.20366 · HTML: https://mlsanity.com/paper/2410.20366/

## Key facts

- Cited by 13 papers in the map (data through 14 Oct 2026).
- Builds on 34 of its 58 references that are in the map.
- First posted on arXiv 27 Oct 2024.

## Builds on (most cited first, 30 of 34)

- [Semi-Supervised Classification with Graph Convolutional Networks](https://mlsanity.com/paper/1609.02907/) (2016)
- [Graph Attention Networks](https://mlsanity.com/paper/1710.10903/) (2017)
- [Inductive Representation Learning on Large Graphs](https://mlsanity.com/paper/1706.02216/) (2017)
- [How Powerful are Graph Neural Networks?](https://mlsanity.com/paper/1810.00826/) (2018)
- [Neural Message Passing for Quantum Chemistry](https://mlsanity.com/paper/1704.01212/) (2017)
- [Deformable Convolutional Networks](https://mlsanity.com/paper/1703.06211/) (2017)
- [Variational Graph Auto-Encoders](https://mlsanity.com/paper/1611.07308/) (2016)
- [Graph Contrastive Learning with Augmentations](https://mlsanity.com/paper/2010.13902/) (2020)
- [Hierarchical Graph Representation Learning with Differentiable Pooling](https://mlsanity.com/paper/1806.08804/) (2018)
- [TUDataset: A collection of benchmark datasets for learning with graphs](https://mlsanity.com/paper/2007.08663/) (2020)
- [Cluster-GCN: An Efficient Algorithm for Training Deep and Large Graph Convolutional Networks](https://mlsanity.com/paper/1905.07953/) (2019)
- [Graph U-Nets](https://mlsanity.com/paper/1905.05178/) (2019)
- [GraphMAE: Self-Supervised Masked Graph Autoencoders](https://mlsanity.com/paper/2205.10803/) (2022)
- [SimGRACE: A Simple Framework for Graph Contrastive Learning without Data Augmentation](https://mlsanity.com/paper/2202.03104/) (2022)
- [GraphMAE2: A Decoding-Enhanced Masked Self-Supervised Graph Learner](https://mlsanity.com/paper/2304.04779/) (2023)
- [Propagation Kernels](https://mlsanity.com/paper/1410.3314/) (2014)
- [Deep Graph-level Anomaly Detection by Glocal Knowledge Distillation](https://mlsanity.com/paper/2112.10063/) (2021)
- [Image Deformation Meta-Networks for One-Shot Learning](https://mlsanity.com/paper/1905.11641/) (2019)
- [A Large-Scale Database for Graph Representation Learning](https://mlsanity.com/paper/2011.07682/) (2020)
- [On Using Classification Datasets to Evaluate Graph-Level Outlier Detection: Peculiar Observations and New Insights](https://mlsanity.com/paper/2012.12931/) (2020)
- [Raising the Bar in Graph-level Anomaly Detection](https://mlsanity.com/paper/2205.13845/) (2022)
- [Diversity-Measurable Anomaly Detection](https://mlsanity.com/paper/2303.05047/) (2023)
- [Towards Self-Interpretable Graph-Level Anomaly Detection](https://mlsanity.com/paper/2310.16520/) (2023)
- [Explainable Classification of Brain Networks via Contrast Subgraphs](https://mlsanity.com/paper/2006.05176/) (2020)
- [Towards Deep Attention in Graph Neural Networks: Problems and Remedies](https://mlsanity.com/paper/2306.02376/) (2023)
- [A Survey on Hypergraph Neural Networks: An In-Depth and Step-By-Step Guide](https://mlsanity.com/paper/2404.01039/) (2024)
- [Directed Graph Auto-Encoders](https://mlsanity.com/paper/2202.12449/) (2022)
- [Graph-level Anomaly Detection via Hierarchical Memory Networks](https://mlsanity.com/paper/2307.00755/) (2023)
- [HypeBoy: Generative Self-Supervised Representation Learning on Hypergraphs](https://mlsanity.com/paper/2404.00638/) (2024)
- [Classification of Edge-dependent Labels of Nodes in Hypergraphs](https://mlsanity.com/paper/2306.03032/) (2023)

## Built on by (most cited first, 13 of 13)

- [FedCIGAR: A Personalized Reconstruction Approach for Federated Graph-level Anomaly Detection](https://mlsanity.com/paper/2605.09428/) (2026)
- [A Survey on Foundation-Model-Based Industrial Defect Detection](https://mlsanity.com/paper/2502.19106/) (2025)
- [Graph Evidential Learning for Anomaly Detection](https://mlsanity.com/paper/2506.00594/) (2025)
- ['Hello, World!': Making GNNs Talk with LLMs](https://mlsanity.com/paper/2505.20742/) (2025)
- [Learnable Kernel Density Estimation for Graphs and Its Application to Graph-Level Anomaly Detection](https://mlsanity.com/paper/2505.21285/) (2025)
- [Learning to Flow from Generative Pretext Tasks for Neural Architecture Encoding](https://mlsanity.com/paper/2510.18360/) (2025)
- [Detect by Yourself: Self-Designing Agentic Workflows for Few-Shot Graph Anomaly Detection](https://mlsanity.com/paper/2605.27470/) (2026)
- [DeNoise: Learning Robust Graph Representations for Unsupervised Graph-Level Anomaly Detection](https://mlsanity.com/paper/2511.04086/) (2025)
- [Physic-HM: Restoring Physical Generative Logic in Multimodal Anomaly Detection via Hierarchical Modulation](https://mlsanity.com/paper/2512.21650/) (2025)
- [Interpretable Graph-Level Anomaly Detection via Contrast with Normal Prototypes](https://mlsanity.com/paper/2602.10708/) (2026)
- [GAD in the Wild: Benchmarking Graph Anomaly Detection under Realistic Deployment Challenges](https://mlsanity.com/paper/2605.07133/) (2026)
- [Cross-Domain Industrial Fault Detection by Causal Mechanism Monitoring](https://mlsanity.com/paper/2608.14666/) (2026)
- [GLASS: Graph-Language Alignment with Spherical Scoring for Transferable Graph-Level Anomaly Detection](https://mlsanity.com/paper/2609.05253/) (2026)

## Abstract (arXiv)

Graph autoencoders (Graph-AEs) learn representations of given graphs by aiming to accurately reconstruct them. A notable application of Graph-AEs is graph-level anomaly detection (GLAD), whose objective is to identify graphs with anomalous topological structures and/or node features compared to the majority of the graph population. Graph-AEs for GLAD regard a graph with a high mean reconstruction error (i.e. mean of errors from all node pairs and/or nodes) as anomalies. Namely, the methods rest on the assumption that they would better reconstruct graphs with similar characteristics to the majority. We, however, report non-trivial counter-examples, a phenomenon we call reconstruction flip, and highlight the limitations of the existing Graph-AE-based GLAD methods. Specifically, we empirically and theoretically investigate when this assumption holds and when it fails. Through our analyses, we further argue that, while the reconstruction errors for a given graph are effective features for GLAD, leveraging the multifaceted summaries of the reconstruction errors, beyond just mean, can further strengthen the features. Thus, we propose a novel and simple GLAD method, named MUSE. The key innovation of MUSE involves taking multifaceted summaries of reconstruction errors as graph features for GLAD. This surprisingly simple method obtains SOTA performance in GLAD, performing best overall among 14 methods across 10 datasets.