An Uncertainty-Aware Pseudo-Label Selection Framework using Regularized Conformal Prediction

Matin Moezzi
8/30/2023
cs.LG

Official code: matinmoezzi/ups_conformal_classification (7 stars, PyTorch, last commit 2021-07-07).

Abstract

Consistency regularization-based methods are prevalent in semi-supervised learning (SSL) algorithms due to their exceptional performance. However, they mainly depend on domain-specific data augmentations, which are not usable in domains where data augmentations are less practicable. On the other hand, Pseudo-labeling (PL) is a general and domain-agnostic SSL approach that, unlike consistency regularization-based methods, does not rely on the domain. PL underperforms due to the erroneous high-confidence predictions from poorly calibrated models. This paper proposes an uncertainty-aware pseudo-label selection framework that employs uncertainty sets yielded by the conformal regularization algorithm to fix the poor calibration neural networks, reducing noisy training data. The codes of this work are available at: https://github.com/matinmoezzi/ups conformal classification

View on arXivView PDF

Code Implementations(1)

matinmoezzi/ups_conformal_classificationOfficial100%
PyTorch

An Uncertainty-Aware Pseudo-Label Selection Framework using Regularized Conformal Prediction

71Python, Jupyter NotebookJul 5, 20215 years ago
conformal-predictionconvolutional-neural-networksdeep-learningimage-classificationpseudo-labeling+3 more

Similar Papers

Complementing Semi-Supervised Learning with Uncertainty Quantification

87% match2022

C-Adapter: Adapting Deep Classifiers for Efficient Conformal Prediction Sets

86% match2 repos2024

Exploring Beyond Logits: Hierarchical Dynamic Labeling Based on Embeddings for Semi-Supervised Classification

85% match2024

Shrinking Class Space for Enhanced Certainty in Semi-Supervised Learning

85% match1 repo482026

NP-Match: When Neural Processes meet Semi-Supervised Learning

85% match1 repo962022

Improving Pseudo-labelling and Enhancing Robustness for Semi-Supervised Domain Generalization

85% match1 repo112024

Cite this paper

@article{moezzi2023uncertaintyaware,
  title  = {An Uncertainty-Aware Pseudo-Label Selection Framework using Regularized Conformal Prediction},
  author = {Matin Moezzi},
  year   = {2023},
  eprint = {2309.15963},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2309.15963v1}
}

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/2309.15963.svg)](https://codeofpaper.com/paper/2309.15963)

Discussion