Research map / Graph learning / Graph self-supervised learning
Semi-Supervised Hierarchical Graph Classification
Jia Li, Yongfeng Huang, Heng Chang, Yu Rong
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 cautious iterative pseudo-label enlargement (SEAL to SEAL-CI) (ablation) | SEAL · Synthetic hierarchical graph | accuracy: 87.8% -> 91.2% | Improved |
| add cautious iterative pseudo-label enlargement (ablation) | SEAL · arXiv paper data | accuracy: 78.8% -> 79.4% | Improved |
| add cautious iterative pseudo-label enlargement (ablation) | SEAL · Tencent QQ group data | Macro-F1: 84.3% -> 85.5% | Improved |
| use hierarchical graph connections (Bert-HC) instead of independent classification (Bert-IC) | Bert features · arXiv paper data | accuracy: 70.1% -> 75.2% | Improved |
| add graph instance structure to Bert features (Bert-IC vs Bert-MLP) | Bert features · arXiv paper data | accuracy: 63.7% -> 70.1% | Improved |
| use HGMI-based SEAL instead of SAGE ignoring hierarchy | SAGE · Tencent QQ group data | Macro-F1: 75.0% -> 84.3% | Improved |
| use SEAL instead of MIRACLE | MIRACLE · Tencent QQ group data | Macro-F1: 80.3% -> 84.3% | Improved |
| use supervised embeddings (SAGE) instead of unsupervised graph2vec | graph2vec-GCN · Synthetic hierarchical graph | accuracy: 85.2% -> 85.7% | Improved |
| increase lambda (number of pseudo-labels) in SEAL-CI (ablation) | SEAL-CI · Tencent QQ group data | false prediction rate: 2.4% | Hurt |
| use kernel methods instead of embedding methods | GK-SVM/WL-SVM · arXiv paper data | accuracy: 38.7% -> 35.1% | Hurt |
Closest papers by meaning
- Hypergraph-enhanced Dual Semi-supervised Graph Classification (ICML 2024 · similarity 0.88)
- Revisiting Semi-Supervised Learning with Graph Embeddings (ICML 2016 · similarity 0.87)
- HGDL: Heterogeneous Graph Label Distribution Learning (NeurIPS 2024 · similarity 0.85)
- Fast Algorithms for Hypergraph PageRank with Applications to Semi-Supervised Learning (ICML 2024 · similarity 0.84)
- DHAKR: Learning Deep Hierarchical Attention-Based Kernelized Representations for Graph Classification (AAAI 2025 · similarity 0.84)
- TGLsta: Low-resource Textual Graph Learning with Semantic and Topological Awareness via LLMs (AAAI 2025 · similarity 0.84)
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.
