• Home
  • Search
  • Feature Space-Preserving Machine Unlearning for Robust Image Classification with Noisy Labels
  • https://doi.org/10.1109/access.2026.3676403Copy DOI Icon

Feature Space-Preserving Machine Unlearning for Robust Image Classification with Noisy Labels

Show More
  • Abstract
  • Literature Map
  • Similar Papers
Abstract

Deep neural networks excel in image classification with clean data, but their performance deteriorates significantly when trained with noisy labels. In contrast to conventional approaches, which address this challenge through sample selection or loss correction, this study proposes a fundamentally different paradigm in which machine unlearning is employed to selectively forget noisy samples. Existing unlearning methods such as scalable remembering and unlearning unbound cause detrimental feature space dispersion, which creates ambiguous decision boundaries. Thus, this paper introduces a feature space-preserving unlearning framework that maintains discriminative structures while removing the influence of noisy labels. The proposed approach computes class centroids from reliable samples and implements a dual optimization strategy in which noisy samples are pushed away from all centroids while clean samples are simultaneously pulled toward their appropriate class centers. Comprehensive experiments on the CIFAR-10 and CIFAR-100 datasets with symmetric and asymmetric noise (ranging from 10%–90%) show that our method consistently outperforms standard unlearning techniques, achieving accuracy improvements of up to 27.61 percentage points on the more complex CIFAR-100 dataset under high-noise conditions. Furthermore, we confirm that our method remains effective even when applied to models trained with state-of-the-art noisy label learning methods. Visualization analysis demonstrates that our approach preserves compact class clusters with clear boundaries, despite aggressive unlearning. The reported implementation is available at https://github.com/meruemon/FSPMU.

Similar Papers
  • Research Article
  • Citations4

Agree to Disagree: Robust Anomaly Detection with Noisy Labels

  • Feb 10, 2025
  • Proceedings of the ACM on Management of Data
  • Dennis M Hofmann +6
  • Research Article

Outlier-trimmed dual-interval smoothing loss for sample selection in learning with noisy labels.

  • Nov 01, 2025
  • Neural networks : the official journal of the International Neural Network Society
  • Senyu Hou +4
  • Research Article
  • Citations9

TC-Net: Detecting Noisy Labels Via Transform Consistency

  • Jan 01, 2022
  • IEEE Transactions on Multimedia
  • Rumeng Yi +1
  • Conference Article

A Robust Learning Framework Using Self-supervised Learning for Learning with Noisy Labels

  • Oct 18, 2022
  • Ryota Higashimoto +2
  • Research Article
  • Citations10

TBC-MI : Suppressing noise labels by maximizing cleaning samples for robust image classification

  • Jun 12, 2024
  • Information Processing and Management
  • Yanhong Li +3
  • Conference Article
  • Citations1

Distribution Mining and Noise Correction for Facial Expression Recognition

  • Apr 19, 2024
  • Hailan Huang +1
  • PDF
  • Research Article
  • Citations3

Separating hard clean samples from noisy samples with samples’ learning risk for DNN when learning with noisy labels

  • Mar 01, 2024
  • Complex & Intelligent Systems
  • Lihui Deng +5
  • PDF
  • Research Article
  • Citations1

An Improved Heteroscedastic Modeling Method for Chest X-ray Image Classification with Noisy Labels

  • May 04, 2023
  • Algorithms
  • Qingji Guan +2
  • Research Article
  • Citations11

Learning with noisy labels using collaborative sample selection and contrastive semi-supervised learning

  • Apr 27, 2024
  • Knowledge-Based Systems
  • Qing Miao +6
  • Research Article

FedEBM: Robust graph federated learning via energy-based model.

  • May 28, 2026
  • Neural networks : the official journal of the International Neural Network Society
  • Jiayu Wang +4
  • Conference Article
  • Citations3

TSS-Net: two-stage with sample selection and semi-supervised net for deep learning with noisy labels

  • Jan 13, 2023
  • Xin Lyu +6
  • Research Article
  • Citations2

Cyclic Learning Rate-Based Co-Training for Image Classification With Noisy Labels

  • Jan 01, 2025
  • IEEE Access
  • Ying Zheng +6
  • PDF
  • Research Article
  • Citations4

CRAS: Curriculum Regularization and Adaptive Semi-Supervised Learning with Noisy Labels

  • Jan 31, 2024
  • Applied Sciences
  • Ryota Higashimoto +2
  • Conference Article
  • Citations33

ProMix: Combating Label Noise via Maximizing Clean Sample Utility

  • Aug 01, 2023
  • Ruixuan Xiao +6
  • Research Article

Alleviating Confirmation Bias in Learning with Noisy Labels via Two-Network Collaboration

  • Mar 11, 2025
  • ACM Transactions on Intelligent Systems and Technology
  • Chenglong Xu +4
Cactus Communications logo

Copyright 2026 Cactus Communications. All rights reserved.