Research map / Graph learning / Knowledge and temporal graphs
ATP: Directed Graph Embedding with Asymmetric Transitivity Preservation
Jiankai Sun, Bortik Bandyopadhyay, Armin Bashizade, Jiongqian Liang, P. Sadayappan, Srinivasan Parthasarathy
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 |
|---|---|---|---|
| Combine graph hierarchy and reachability in ATP's objective matrix and factorize it with NMF rather than SVD | Directed graph embedding · Link prediction, four datasets | AUC: ATP improves over SVDM by $22.01\%$ among all datasets | Improved |
| Use the ATP hierarchy-and-reachability matrix instead of HOPE's higher-order proximity matrices | Directed graph embedding · Link prediction, four datasets | AUC: ATP improves over HOPE-KI from $19.84\%$ to $51.97\% | Improved |
| Use a nonlinear log transformation instead of the constant adjacency matrix (ablation) | ATP · Link prediction, four datasets | AUC: ATP-log improves over ATP-Constant by $8.51\%$ in average among all datasets | Improved |
| Use a nonlinear log transformation instead of a linear transformation (ablation) | ATP · GNU | AUC: ATP-log improves over ATP-Linear by $4.74\%$ on the largest dataset GNU | Improved |
| Use harmonic transformation instead of log transformation (ablation) | ATP · Link prediction, four datasets | AUC | No clear effect |
| Use graph reachability from transitive closure rather than first-order neighbors alone (ablation) | Hierarchy-based embedding · Cit-HepPH | AUC: SA-linear improves over SA-linear-Adj by $3.1\%$ on Cit-HepPH | Improved |
| Apply log transformation rather than linear transformation (ablation) | Hierarchy-and-reachability embedding · GNU | AUC: SA-log-Adj improves over SA-linear-Adj by $4.31\%$ on GNU | Improved |
| Use ATP for question difficulty estimation rather than TrueSkill | Question difficulty estimation · Five Stack Exchange sites | Pairwise accuracy: ATP improves over TrueSkill ... on average by $6.56\% | Improved |
| Use ATP for question difficulty estimation rather than Number-of-Answers | Question difficulty estimation · Five Stack Exchange sites | Pairwise accuracy: ATP improves over ... Number-of-Answer ... on average by $5.92\% | Improved |
| Use ATP for question difficulty estimation rather than QDEE | Question difficulty estimation · Five Stack Exchange sites | Pairwise accuracy: ATP improves over ... QDEE on average by $5.06\% | Improved |
| Use ATP for cold question routing rather than ColdRoute | Cold question routing · Five Stack Exchange sites | Accuracy: ATP improves upon routing metric Accuracy over ColdRoute by $6.14\% | Improved |
| Use ATP for cold question routing rather than QDEE | Cold question routing · Five Stack Exchange sites | MRR: ATP improves upon routing metric MRR over QDEE by $7.59\% | Improved |
Closest papers by meaning
- G-Retriever: Retrieval-Augmented Generation for Textual Graph Understanding and Question Answering (NeurIPS 2024 · similarity 0.87)
- Personalized Question Routing via Heterogeneous Network Embedding (AAAI 2019 · similarity 0.86)
- KG-Guard: Graph-Based Hallucination Detection for Knowledge Base Question Answering (NeurIPS 2026 · similarity 0.85)
- A General Framework for Producing Interpretable Semantic Text Embeddings (ICLR 2025 · similarity 0.84)
- Harnessing Large Language Models for Knowledge Graph Question Answering via Adaptive Multi-Aspect Retrieval-Augmentation (AAAI 2025 · similarity 0.84)
- Debate on Graph: A Flexible and Reliable Reasoning Framework for Large Language Models (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.
