Descent-to-Delete: Gradient-Based Methods for Machine Unlearning

Abstract

We study the data deletion problem for convex models. By leveraging techniques from convex optimization and reservoir sampling, we give the first data deletion algorithms that are able to handle an arbitrarily long sequence of adversarial updates while promising both per-deletion run-time and steady-state error that do not grow with the length of the update sequence. We also introduce several new conceptual distinctions: for example, we can ask that after a deletion, the entire state maintained by the optimization algorithm is statistically indistinguishable from the state that would have resulted had we retrained, or we can ask for the weaker condition that only the observable output is statistically indistinguishable from the observable output that would have resulted from retraining. We are able to give more efficient deletion algorithms under this weaker deletion criterion.

References in corpus (4)

• Making AI Forget You: Data Deletion in Machine Learning
• Comunication-Efficient Algorithms for Statistical Optimization
• Certified Data Removal from Machine Learning Models
• Pan-Private Uniformity Testing

Cited by in corpus (11)

• Fast Yet Effective Machine Unlearning
• When Machine Unlearning Jeopardizes Privacy
• Zero-Shot Machine Unlearning
• GIF: A General Graph Unlearning Strategy via Influence Function
• Towards Probabilistic Verification of Machine Unlearning
• Coded Machine Unlearning
• Landing AI on Networks: An equipment vendor viewpoint on Autonomous Driving Networks
• Certifiable Machine Unlearning for Linear Models
• Machine Unlearning via Algorithmic Stability
• Federated Unlearning
• Lightweight machine unlearning in neural network