Research map / Learning theory and algorithms / Prediction and decision losses
Beyond Imputation: Mask-Adaptive Conformal Prediction via Tree Embeddings on General Missing Data Mechanisms
Jiarong Fan, Juhyun Park, Thi Phuong Thuy Vo, Nicolas Brunel
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes a mask-adaptive conformal prediction method using tree embeddings to deliver coverage guarantees across missingness patterns under general missing-data mechanisms.
Problem
Conformal prediction with missing values: marginal coverage can hide systematic undercoverage for certain missingness patterns, and existing conditional-coverage methods often fail when values are missing.
Approach
Mask-adaptive conformal prediction that uses tree embeddings to adapt to missingness patterns, avoiding reliance on imputation and handling general missing data mechanisms.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use tree embedding instead of MICE-imputed covariates (both with raw mask) (ablation) | mask-adaptive CP · Synthetic PMIX | coverage: 0.8476 -> 0.9145 | Improved |
| use tree embedding instead of MICE-imputed covariates (ablation) | mask-adaptive CP · MIMIC-IV LOS | shortfall risk | Improved |
| add tree embeddings (vs mask-only linear representation, zero trees) (ablation) | mask-adaptive CP · Synthetic MAR d=6 | coverage | Improved |
| increase number of trees beyond 10 (ablation) | mask-adaptive CP · Synthetic MAR d=6 | coverage and length | No clear effect |
| replace Exact/Nested/Weighted/ARC baselines with mask-adaptive CP | mask-adaptive CP · Synthetic PMIX d=3 | interval length | Improved |
| use mask-adaptive CP vs baselines | mask-adaptive CP · Concrete (self-masked MNAR) | shortfall risk | Improved |
| use mask-adaptive CP vs baselines | mask-adaptive CP · Concrete | interval length | Hurt |
| use mask-adaptive CP vs two remaining baselines | mask-adaptive CP · MIMIC-IV LOS | shortfall risk: average shortfall below 0.5% | Hurt |
| use mask-adaptive CP vs baselines in MCAR and MAR | mask-adaptive CP · Synthetic d=3 | shortfall risk and length | No clear effect |
| use mask-adaptive CP on rare masks (fewer than 10 calibration occurrences) | mask-adaptive CP · Synthetic | coverage | No clear effect |
Closest papers by meaning
- Learning What Not to Impute: An Uncertainty-Aware Diffusion Framework for Meaningful Missingness (NeurIPS 2026 · similarity 0.89)
- Rectifying Conformity Scores for Better Conditional Coverage (ICML 2025 · similarity 0.87)
- SimplexUQ: An Evaluation Framework and Benchmark for Conformal Uncertainty on Simplex-Valued Predictions (NeurIPS 2026 · similarity 0.87)
- Generative Conformal Prediction with Optimized Coverage Allocation (NeurIPS 2026 · similarity 0.87)
- Prediction models that learn to avoid missing values (ICML 2025 · similarity 0.87)
- Conformal Prediction with Cellwise Outliers: A Detect-then-Impute Approach (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.
