analiza aiqbench
This research presents QpiGNN, a novel Graph Neural Network framework designed for efficient uncertainty quantification without relying on costly resampling or post-hoc calibration. By directly optimizing coverage through a dual-head architecture trained with label-only supervision, the method achieves significantly narrower prediction intervals and robust performance across diverse benchmarks. The work advances machine learning reliability in high-stakes applications where traditional exchangeability assumptions often fail in graph-structured data.
Quantile-Free Uncertainty Quantification in Graph Neural Networks
arXiv:2605.04847v2 Announce Type: replace-cross Abstract: Uncertainty quantification (UQ) in graph neural networks (GNNs) is crucial in high-stakes domains but remains a significant challenge. In graph settings, message passing often relies on strong assumptions such as exchangeability, which are rarely satisfied in practice, and achieving reliable UQ typically requires costly resampling or post-hoc calibration. To address these issues, we introduce Quantile-free Prediction Interval GNN (QpiGNN), a framework that builds on quantile regression (QR) to enable GNN-based UQ by directly optimizing coverage and interval width without requiring quantile inputs or post-processing. QpiGNN employs a dual-head architecture that decouples prediction and uncertainty, and is trained with label-only supervision through a quantile-free joint loss. This design allows efficient training and yields robust prediction intervals, with theoretical guarantees of asymptotic coverage and near-optimal width under mild assumptions. Experiments on 19 synthetic and real-world benchmarks show QpiGNN achieves average 22% higher coverage and 50% narrower intervals than baselines, while ensuring efficiency and robustness to noise and structural shifts.
Abstract:Uncertainty quantification (UQ) in graph neural networks (GNNs) is crucial in high-stakes domains but remains a significant challenge. In graph settings, message passing often relies on strong assumptions such as exchangeability, which are rarely satisfied in practice, and achieving reliable UQ typically requires costly resampling or post-hoc calibration. To address these issues, we introduce Quantile-free Prediction Interval GNN (QpiGNN), a framework that builds on quantile regression (QR) to enable GNN-based UQ by directly optimizing coverage and interval width without requiring quantile inputs or post-processing. QpiGNN employs a dual-head architecture that decouples prediction and uncertainty, and is trained with label-only supervision through a quantile-free joint loss. This design allows efficient training and yields robust prediction intervals, with theoretical guarantees of asymptotic coverage and near-optimal width under mild assumptions. Experiments on 19 synthetic and real-world benchmarks show QpiGNN achieves average 22% higher coverage and 50% narrower intervals than baselines, while ensuring efficiency and robustness to noise and structural shifts.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2605.04847 [cs.LG]
(or arXiv:2605.04847v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2605.04847 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 252 cuvinte
Subiecte conexe
In-Context Residual Calibration for Uncertainty Quantification of Energy Time Series over Graphs
arXiv · 8 oct. 2026
HeAD-CP: Heterophily-Aware Diffused Conformal Prediction Sets for Graph Neural Networks
arXiv · 29 iul. 2026
Conformal Prediction for Time Series with Deep Sequence Models
arXiv · 5 oct. 2026
Interval and fuzzy physics-augmented neural networks (iPANN and fPANN) for uncertainty quantification and propagation in constitutive modeling
arXiv · 23 iul. 2026