Label Perturbation

Label perturbation, the intentional or unintentional alteration of training data labels, is a significant area of research impacting various machine learning models. Current research focuses on understanding the effects of label noise on model performance and robustness, exploring techniques like adaptive label perturbation to improve model calibration and mitigate overfitting, and investigating the impact of label perturbations on explainability in models like graph neural networks and large language models. These studies are crucial for developing more reliable and trustworthy machine learning systems, particularly in safety-critical applications, and for improving our fundamental understanding of how these models learn and generalize.

Papers

August 26, 2024

Theoretical Proportion Label Perturbation for Learning from Label Proportions in Large Bags
Shunsuke Kubo, Shinnosuke Matsuo, Daiki Suehiro, Kazuhiro Terada, Hiroaki Ito, Akihiko Yoshizawa, Ryoma Bise
LeArning Abstract Noisy Label Weakly Supervised Learning Instance Level Bag Prototype Label Proportion Label Perturbation

September 4, 2023

On the Robustness of Post-hoc GNN Explainers to Label Noise
Zhiqiang Zhong, Yangqianzi Jiang, Davide Mottin
Native Robustness Label Noise GNN Explanation Effective Explanation GNN Performance Label Perturbation

July 28, 2023

Investigating the Learning Behaviour of In-context Learning: A Comparison with Supervised Learning
Xindi Wang, Yufei Wang, Can Xu, Xiubo Geng, Bowen Zhang, Chongyang Tao, Frank Rudzicz, Robert E. Mercer, Daxin Jiang
Large Language Model Context Learning Supervised Learning Consistent Comparison Behaviour Learning Downstream Task Performance Label Perturbation

July 25, 2023

Model Calibration in Dense Classification with Adaptive Label Perturbation
Jiawei Liu, Changkun Ye, Shan Wang, Ruikai Cui, Jing Zhang, Kaihao Zhang, Nick Barnes
Model Calibration Dense Training Binary Cross Entropy Label Perturbation

September 13, 2022

Class-Level Logit Perturbation
Mengyang Li, Fengguang Su, Ou Wu, Ji Zhang
Feature Perturbation Label Perturbation Negative Data Augmentation Logit Vector

August 10, 2022

Customized Watermarking for Deep Neural Networks via Label Distribution Perturbation
Tzu-Yun Chien, Chih-Ya Shen
Deep Neural Network Watermarking Method Text Watermarking Watermark Removal Attack Label Perturbation Deep Neural Network Watermarking

July 11, 2022

Towards Effective Multi-Label Recognition Attacks via Knowledge Graph Consistency
Hassan Mahmood, Ehsan Elhamifar
Adversarial Attack Multi Label Knowledge Alignment Label Attack Label Perturbation

May 23, 2022

Representation Projection Invariance Mitigates Representation Collapse
Anastasia Razdaibiedina, Ashish Khetan, Zohar Karnin, Daniel Khashabi, Vishaal Kapoor, Vivek Madan
Pre Trained Language Model Representation Collapse Label Perturbation