探索增量式遗忘：技术、挑战与未来方向
摘要
随着数据隐私在机器学习（Machine Learning, ML）应用中日益增长的需求，机器学习遗忘（Machine Unlearning, MU）已成为 critical 研究领域。随着“被遗忘权”在全球范围内被规范化，越来越重要的是开发能够在保持系统性能和可扩展性的同时删除用户数据的机制。增量式遗忘（Incremental Unlearning, IU）是一种有前景的 MU 解决方案，用于高效地从 ML 模型中删除特定数据，而无需进行昂贵且耗时的完整重新训练。本文介绍了 various techniques and approaches to IU. It explores the challenges faced in designing and implementing IU mechanisms. Datasets and metrics for evaluating the performance of unlearning techniques are discussed as well. Finally, potential solutions to the IU challenges alongside future research directions are offered. This survey provides valuable insights for researchers and practitioners seeking to understand the current landscape of IU and its potential for enhancing privacy-preserving intelligent systems.
引用
@article{arxiv.2502.16708,
title = {Exploring Incremental Unlearning: Techniques, Challenges, and Future Directions},
author = {Sadia Qureshi and Thanveer Shaik and Xiaohui Tao and Haoran Xie and Lin Li and Jianming Yong and Xiaohua Jia},
journal= {arXiv preprint arXiv:2502.16708},
year = {2025}
}
备注
This work has been submitted to the IEEE for possible publication