Research map / Learning theory and algorithms / Prediction and decision losses
Zono-Conformal Prediction: Zonotope-Based Uncertainty Quantification for Regression and Classification Tasks
Laura Lützow, Michael Eichelbeck, Mykel J. Kochenderfer, Matthias Althoff
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 zonotope prediction sets instead of axis-aligned intervals (ZCP vs IPM and CP) | zono-conformal prediction · regression and classification tasks | conservatism (prediction set volume or classes) | Improved |
| zonotopes on tasks with correlated outputs (SD-R2, Energy, Photovoltaic) | zono-conformal prediction · SD-R2, Energy, Photovoltaic | prediction set size | Improved |
| zonotopes when selected uncertainties fail to capture correlations (SD-R1, BlogPost) | zono-conformal prediction · SD-R1, BlogPost | conservatism | No clear effect |
| use ZCP instead of CP in terms of coverage guarantee | zono-conformal prediction · regression, epsilon for 0.9 guarantee | smallest epsilon: Energy n_out=0: 0.0496 -> 0.2508 | Hurt |
| use ZCP instead of IPM or CP on classification tasks | zono-conformal prediction · MNIST and Covertype | number of predicted classes and coverage | Mixed |
| select all output uncertainties by default (ORand, ORand*, QR instead of Rand) (ablation) | uncertainty placement · SD-R1, SD-R2, Energy, SCM20D | normalized conservatism | Improved |
| replace QR uncertainty placement with simple ORand (ablation) | uncertainty placement · SD-R1, SD-R2, Energy, SCM20D | prediction set size | Improved |
| add random generators to the generator template (ORand*) (ablation) | uncertainty placement · SD-R1, SD-R2, Energy, SCM20D | conservatism and computation time | Mixed |
| identify more uncertainties (increase p_p) (ablation) | uncertainty placement · SD-R1, SD-R2, Energy, SCM20D | conservatism and computation time | Mixed |
| use Rotated Intervals cost instead of Interval cost (ablation) | cost function · regression data sets | prediction set volume: reduction by a factor of 0.66 | Improved |
| increase number of rotations n_r in RI cost (ablation) | cost function · regression data sets | conservatism and computation time: SCM20D time: 0.81 -> 1.03 | Mixed |
| use Score or Score Difference cost instead of Interval/RI for classification (ablation) | cost function · classification data sets | normalized conservatism | No clear effect |
Closest papers by meaning
- Self-Calibrating Conformal Prediction (NeurIPS 2024 · similarity 0.87)
- Optimal transport-based conformal prediction (ICML 2025 · similarity 0.87)
- Conformal Prediction under Lévy-Prokhorov Distribution Shifts: Robustness to Local and Global Perturbations (NeurIPS 2025 · similarity 0.87)
- Generative Conformal Prediction with Optimized Coverage Allocation (NeurIPS 2026 · similarity 0.87)
- Direct Prediction Set Minimization via Bilevel Conformal Classifier Training (ICML 2025 · similarity 0.86)
- Conformal Prediction as Bayesian Quadrature (ICML 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.
