Research map / Graph learning / GNN architectures
Quantile-Free Uncertainty Quantification in Graph Neural Networks
Soyoung Park, Hwanjun Song, Sungsu Lim
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 |
|---|---|---|---|
| use quantile-free joint loss instead of RQR loss (ablation) | QpiGNN · BA (synthetic) | PICP: 0.19 -> 0.98 | Improved |
| use quantile-free joint loss instead of RQR loss (ablation) | QpiGNN · Tree (synthetic) | PICP: 0.27 -> 0.96 | Improved |
| use quantile-free joint loss instead of RQR loss (ablation) | QpiGNN · Chameleon | MPIW: 0.30 -> 0.23 | Improved |
| lower width penalty lambda from 0.5 to 0.1 | QpiGNN · BA (synthetic) | MPIW: 0.49 -> 1.01 | Hurt |
| use Bayesian-optimized lambda instead of 0.5 | QpiGNN · Chameleon | PICP: 0.51 -> 0.96 | Improved |
| use Bayesian-optimized lambda instead of 0.5 | QpiGNN · Twitch | PICP: 0.59 -> 0.94 | Improved |
| add dual-head architecture with learnable margin (ablation) | QpiGNN · nine synthetic datasets | PICP-MPIW trade-off | Improved |
| add coverage and width terms jointly in loss (ablation) | QpiGNN · nine synthetic datasets | PICP-MPIW trade-off | Improved |
| increase edge dropout from 0.2 to 0.6 | QpiGNN · ER graphs | PICP: 0.84 -> 0.91 | Improved |
| increase target noise sigma from 0.1 to 0.3 | QpiGNN · ER graphs | MPIW: 0.44 -> 1.05 | Hurt |
| use community split instead of random split | QpiGNN · Education | MPIW: 0.90 -> 0.54 | Improved |
| use degree split instead of random split | QpiGNN · Twitch | MPIW: 0.54 -> 0.68 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CF-GNN — CFGNN: Cross Flow Graph Neural Networks for Question Answering on Complex Tables (AAAI 2020) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Accurate and Scalable Estimation of Epistemic Uncertainty for Graph Neural Networks (ICLR 2024 · similarity 0.91)
- A Unified Uncertainty Representation for Graph Neural Networks via Doubly-Spectral Stochastic Expansion (NeurIPS 2026 · similarity 0.89)
- Energy-based Epistemic Uncertainty for Graph Neural Networks (NeurIPS 2024 · similarity 0.88)
- Uncertainty Estimation on Graphs with Structure Informed Stochastic Partial Differential Equations (NeurIPS 2025 · similarity 0.88)
- Uncertainty Estimation for Heterophilic Graphs Through the Lens of Information Theory (ICML 2025 · similarity 0.87)
- Uncertainty Modeling in Graph Neural Networks via Stochastic Differential Equations (ICLR 2025 · similarity 0.87)
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.
