Corrective Machine Unlearning
This addresses data integrity challenges for model developers using large-scale datasets from the Internet, offering a new strategy to handle manipulated data, though it is incremental as it builds on existing unlearning methods.
The paper tackles the problem of mitigating the impact of manipulated or incorrect data in trained machine learning models when only a small subset of the corrupted data is identified, formalizing this as Corrective Machine Unlearning. It finds that most existing unlearning methods require most manipulated data to be identified for effectiveness, but Selective Synaptic Dampening achieves limited success with just a small portion of manipulated samples.
Machine Learning models increasingly face data integrity challenges due to the use of large-scale training datasets drawn from the Internet. We study what model developers can do if they detect that some data was manipulated or incorrect. Such manipulated data can cause adverse effects including vulnerability to backdoored samples, systemic biases, and reduced accuracy on certain input domains. Realistically, all manipulated training samples cannot be identified, and only a small, representative subset of the affected data can be flagged. We formalize Corrective Machine Unlearning as the problem of mitigating the impact of data affected by unknown manipulations on a trained model, only having identified a subset of the corrupted data. We demonstrate that the problem of corrective unlearning has significantly different requirements from traditional privacy-oriented unlearning. We find most existing unlearning methods, including retraining-from-scratch without the deletion set, require most of the manipulated data to be identified for effective corrective unlearning. However, one approach, Selective Synaptic Dampening, achieves limited success, unlearning adverse effects with just a small portion of the manipulated samples in our setting, which shows encouraging signs for future progress. We hope our work spurs research towards developing better methods for corrective unlearning and offers practitioners a new strategy to handle data integrity challenges arising from web-scale training. Code is available at https://github.com/drimpossible/corrective-unlearning-bench.