Conformal Prediction with Partially Labeled Data

Alireza Javanmardi, Yusuf Sale, Paul Hofman, Eyke Hüllermeier
6/1/2023
cs.LGstat.ML

No official code found. Best implementation: pwhofman/conformal-partial-labels (7 stars, last commit 2023-09-20).

Abstract

While the predictions produced by conformal prediction are set-valued, the data used for training and calibration is supposed to be precise. In the setting of superset learning or learning from partial labels, a variant of weakly supervised learning, it is exactly the other way around: training data is possibly imprecise (set-valued), but the model induced from this data yields precise predictions. In this paper, we combine the two settings by making conformal prediction amenable to set-valued training data. We propose a generalization of the conformal prediction procedure that can be applied to set-valued training and calibration data. We prove the validity of the proposed method and present experimental studies in which it compares favorably to natural baselines.

View on arXivView PDF

Code Implementations(1)

pwhofman/conformal-partial-labelsCommunity100%

Conformal prediction for partial label learning

71Python, Jupyter NotebookMay 23, 20233 years ago

Similar Papers

Cost-Sensitive Conformal Training with Provably Controllable Learning Bounds

89% match2025

Conformal Prediction with Learned Features

86% match2024

Split Conformal Classification with Unsupervised Calibration

85% match1 repo12025

Conformal Prediction of Classifiers with Many Classes based on Noisy Labels

85% match1 repo32025

Improving Adaptive Conformal Prediction Using Self-Supervised Learning

85% match2 repos172023

Robust Yet Efficient Conformal Prediction Sets

84% match1 repo12024

Cite this paper

@article{javanmardi2023conformal,
  title  = {Conformal Prediction with Partially Labeled Data},
  author = {Alireza Javanmardi and Yusuf Sale and Paul Hofman and Eyke Hüllermeier},
  year   = {2023},
  eprint = {2306.01191},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2306.01191v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2306.01191.svg)](https://codeofpaper.com/paper/2306.01191)

Discussion