ACD-U: Asymmetric co-teaching with machine unlearning for robust learning with noisy labels
Submitted to ELSEVIER Image and Vision Computing
This repository contains the implementation of ACD-U, a novel framework for learning with noisy labels. The proposed method enables post-hoc correction of overfitted noisy samples and efficient learning for trained models by utilizing a selective forgetting mechanism for overfitted noisy samples and an asymmetric co-learning architecture that leverages different learning characteristics between architectures.
The paper is currently under review at ELSEVIER Image and Vision Computing.
Note:
As the manuscript is under peer review, the repository is currently in a limited-release state. Some details, including datasets, trained models, and complete documentation, will be provided after the review process concludes.
A BibTeX entry will be provided here upon acceptance.