Summary of Detecting Anomalies in Dynamic Graphs Via Memory Enhanced Normality, by Jie Liu et al.
Detecting Anomalies in Dynamic Graphs via Memory enhanced Normality
by Jie Liu, Xuequn Shang, Xiaolin Han, Kai Zheng, Hongzhi Yin
First submitted to arxiv on: 14 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The novel Spatial-Temporal memories-enhanced graph autoencoder (STRIPE) is introduced to tackle the challenge of anomaly detection in dynamic graphs, which typically employ unsupervised learning frameworks. STRIPE initially extracts spatial and temporal features using Graph Neural Networks (GNNs) and gated temporal convolution layers, then incorporates separate memory networks to capture and store prototypes of normal patterns. The stored patterns are retrieved and integrated with encoded graph embeddings through a mutual attention mechanism before being fed into the decoder for reconstruction. This approach minimizes reconstruction errors while emphasizing compactness and distinctiveness of the embeddings. Extensive experiments on six benchmark datasets demonstrate STRIPE’s effectiveness and efficiency, outperforming existing methods by 5.8% in AUC scores and reducing training time by 4.62X. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Anomaly detection in dynamic graphs is a big challenge because the graph changes over time. Usually, this problem is solved using unsupervised learning, but it has some drawbacks. Some methods only focus on spatial normal patterns or ignore temporal patterns. Other methods use contrastive learning with negative sampling, which can be slow and not work well for large graphs. To solve these problems, a new method called STRIPE was developed. STRIPE uses Graph Neural Networks (GNNs) to extract features from the graph, then remembers normal patterns in the graph. It also retrieves and combines this information with other data to detect anomalies. This approach works better than others and is faster too. |
Keywords
* Artificial intelligence * Anomaly detection * Attention * Auc * Autoencoder * Decoder * Unsupervised
