1: \begin{abstract}

2: In the field of machine unlearning, certified unlearning has been extensively studied in convex machine learning models due to its high efficiency and strong theoretical guarantees.

3: However, its application to deep neural networks (DNNs), known for their highly nonconvex nature, still poses challenges.

4: To bridge the gap between certified unlearning and DNNs, 

5: % we study the problem of certified unlearning and propose several simple techniques to extend certified unlearning methods to non-convex objectives.

6: we propose several simple techniques to extend certified unlearning methods to nonconvex objectives.

7: % Additionally, we integrate an inverse Hessian approximation technique into our certified unlearning framework to enhance efficiency without compromising certification guarantees.

8: To reduce the time complexity, we develop an efficient computation method by inverse Hessian approximation without compromising certification guarantees.

9: In addition, we extend our discussion of certification to nonconvergence training and sequential unlearning, considering that real-world users can send unlearning requests at different time points.

10: % Furthermore, considering that users in the real world can send unlearning requests at different time points, we develop the sequential certified unlearning to unlearn different groups of user data sequentially.

11: Extensive experiments on three real-world datasets demonstrate the efficacy of our method and the advantages of certified unlearning in DNNs.

12: \end{abstract}

13: