arXiv：cs.LG· Virgile Dine, Teddy Furon·· 7 小时前AI 评分33
基于代理的机器遗忘（Proxy-Based Unlearning）行为保证
Behavioral Guarantees for Proxy-Based Unlearning
AI 导读
该论文提出一个统一框架，将近期基于代理的遗忘方法推广，并证明遗忘后模型与保留数据理想后验分布之间 KL 散度的上界。方法将近似遗忘建模为约束优化问题，通过在输出空间引入按代理自适应缩放的遗忘信号来保证行为上界，必要时以目标模型为教师将更新蒸馏进权重。在两个遗忘场景实验中，该方法所得分类器最接近从头重训练的模型。
正文
Abstract:This paper proposes a framework generalizing recent proxy-based unlearning methods and proves theoretical guarantees about the behavior of the resulting unlearned model: upper bounds on its Kullback-Leibler divergence to the ideal posterior distribution of the retain data. We model approximate unlearning as a constrained optimization problem and interpret a family of solutions as introducing a scaled unlearning signal in the output space. The unlearning signal arises from proxies of the posterior data distributions. Its scale is adapted to the proxies to ensure the behavioral upper bounds. This framework relies on the structure of the data distributions in order to create proxies. If need be, the target serves as a teacher to distill the update in the weights. Our approach is experimentally validated over two forgetting scenarios as reaching the closest classifier to the model retrained from scratch.
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2605.10680 [cs.LG] |
| (or arXiv:2605.10680v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2605.10680 arXiv-issued DOI via DataCite |
Submission history
From: Virgile Dine [view email]
[v1]
Mon, 11 May 2026 14:57:31 UTC (273 KB)
[v2]
Tue, 6 Oct 2026 16:29:49 UTC (498 KB)
来源：arXiv：cs.LG · arxiv.org