AI Unlearning: From Data Deletion to Verifiable Model Forgetting
中文摘要
本文深入探讨机器遗忘技术，涵盖SISA、梯度遗忘与验证方法，研究如何让AI模型安全、可验证地删除特定训练数据。
English Summary
This article explores machine unlearning techniques like SISA and gradient-based forgetting, detailing how to achieve verifiable data deletion and ensure models effectively forget specific information.
原文节选
A technical deep dive into machine unlearning, gradient-based forgetting, SISA, approximate unlearning, verification, data provenance, and… Continue reading on Medium »