Why Forget-Only Unlearning Needs Memorization
In this work, we study forget-only unlearning, where the deletion algorithm receives only the trained model and the examples to forget, with no retained data or extra training information.
ProofPaper ↗
Key points
- Machine unlearning asks for a deletion algorithm whose output is close to retraining from scratch without the selected forget examples.
- We ask whether forget-only unlearning is always possible.
- Using this observation, we derive lower bounds on how accurately unlearning can match retraining and instantiate them for several standard learning algorithms.
- Overall, our results show that information discarded during ordinary learning may be needed later for deletion, so models designed for forget-only unlearning may need to retain more information than standard training does.
Sources (1)
- [1]Why Forget-Only Unlearning Needs MemorizationarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 05:55 PM
In this work, we study forget-only unlearning, where the deletion algorithm receives only the trained model and the examples to forget, with no retained data or extra training information.
Machine unlearning asks for a deletion algorithm whose output is close to retraining from scratch without the selected forget examples.
Extractive summary: sentences quoted from the sources.