SEGO is an unsupervised framework designed to improve the reliability of graph neural networks (GNNs) by detecting out-of-distribution (OOD) samples during testing. With the increasing amount of unlabeled data, OOD detection is crucial for ensuring the reliability of GNNs. Existing methods often struggle with redundant information in graph structures, impairing their ability to differentiate between in-distribution (ID) and OOD data. SEGO addresses this challenge by integrating structural entropy into OOD detection for graph classification. It introduces an anchor view in the form of a coding tree by minimizing structural entropy, effectively removing redundant information while preserving essential structural information. SEGO also presents a multi-grained contrastive learning scheme at local, global, and tree levels using triplet views. Extensive experiments on real-world datasets demonstrate SEGO's superior performance over state-of-the-art baselines in OOD detection.
Contrastive learning, structural entropy
Graph Neural Networks
Real-world graph datasets
OOD detection accuracy, structural entropy
Research environments
No
Yes
Unsupervised OOD detection, structural entropy
No
Standard computing resources
Linux, Windows, macOS
Compatible with GNN frameworks
N/A
N/A
N/A
Yes
Research community
N/A
Varies based on dataset
Depends on model complexity
Standard for GNNs
N/A
N/A
Focus on specific OOD tasks
AI research, graph analysis
Improving GNN reliability
Researchers
Integrates with GNN frameworks
Scalable with model size
Community support
N/A
Command-line
No
N/A
Open-source
Yes
Research institutions
N/A
N/A
N/A
Research tool
No
N/A
Open-source
0.00
N/A
Open-source
01/01/1970
01/01/1970
N/A
N/A
Yes