Research map / Graph learning / Knowledge and temporal graphs
ESSEN: Improving Evolution State Estimation for Temporal Networks using Von Neumann Entropy
Qiyao Huang, Yingyue Zhang, Zhihong Zhang, Edwin Hancock
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 |
|---|---|---|---|
| remove von Neumann edge entropy bias from attention (ESSEN-E) (ablation) | ESSEN · MathOverflow and BitcoinAlpha link prediction | AUC | Hurt |
| remove virtual evolution node representation (ESSEN-V) (ablation) | ESSEN · MathOverflow and BitcoinAlpha link prediction | AUC | Hurt |
| replace MoTE decoder with simple MLP decoder (ESSEN-D) (ablation) | ESSEN · MathOverflow and BitcoinAlpha link prediction | AUC | Hurt |
| use ESSEN instead of NeurTWs (transductive) | ESSEN · MathOverflow | AUC (%): 92.56 -> 98.60 | Improved |
| use ESSEN instead of NeurTWs (inductive) | ESSEN · MathOverflow | AUC (%): 91.83 -> 98.33 | Improved |
| use ESSEN instead of CAW on Wikipedia transductive | ESSEN · Wikipedia transductive | AUC (%): 98.96 -> 99.03 | Improved |
| use ESSEN instead of CAW on Wikipedia inductive | ESSEN · Wikipedia inductive | AUC (%): 98.75 -> 98.80 | Improved |
| reduce evolution-aware neighborhood size N from 200 to 50 (ablation) | ESSEN · MathOverflow | AUC: 98.56% to 95.43% | Hurt |
| reduce training time split point Ttr (fewer training samples) (ablation) | ESSEN · MathOverflow and BitcoinAlpha inductive | AUC | No clear effect |
| vary number of experts Y in MoTE decoder (ablation) | ESSEN · MathOverflow and BitcoinAlpha | AUC | Mixed |
| cache approximate thermodynamic quantities after first epoch | ESSEN · training time vs NeurTWs | running time | Improved |
Closest papers by meaning
- Temporal Generalization Estimation in Evolving Graphs (ICLR 2024 · similarity 0.86)
- Analytical Construction on Geometric Architectures: Transitioning from Static to Temporal Link Prediction (ICML 2025 · similarity 0.86)
- TMetaNet: Topological Meta-Learning Framework for Dynamic Link Prediction (ICML 2025 · similarity 0.86)
- Using Time-Aware Graph Neural Networks to Predict Temporal Centralities in Dynamic Graphs (NeurIPS 2024 · similarity 0.86)
- State Space Models on Temporal Graphs: A First-Principles Study (NeurIPS 2024 · similarity 0.86)
- DyG-Mamba: Continuous State Space Modeling on Dynamic Graphs (NeurIPS 2025 · similarity 0.86)
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.
