Aviraj (Avi) Newatia
@projectavi.bsky.social
Machine Learning Researcher. Undergraduate Computer Scientist. University of Toronto & Vector Institute.
6/6
In conclusion, RELOAD is an effective algorithm for unlearning arbitrary parts of the training set, and provides strong privacy guarantees for forgotten data.
In conclusion, RELOAD is an effective algorithm for unlearning arbitrary parts of the training set, and provides strong privacy guarantees for forgotten data.
December 14, 2024 at 10:01 PM
6/6
In conclusion, RELOAD is an effective algorithm for unlearning arbitrary parts of the training set, and provides strong privacy guarantees for forgotten data.
In conclusion, RELOAD is an effective algorithm for unlearning arbitrary parts of the training set, and provides strong privacy guarantees for forgotten data.
5/6
Using TabNet attention masks we show how RELOAD removes dependence of model inference on forgotten features.
Using TabNet attention masks we show how RELOAD removes dependence of model inference on forgotten features.
December 14, 2024 at 10:01 PM
5/6
Using TabNet attention masks we show how RELOAD removes dependence of model inference on forgotten features.
Using TabNet attention masks we show how RELOAD removes dependence of model inference on forgotten features.
4/6
We conducted experiments on forgetting random samples and entire features from the training set, consistently outperforming unlearning baselines and protecting user privacy.
We conducted experiments on forgetting random samples and entire features from the training set, consistently outperforming unlearning baselines and protecting user privacy.
December 14, 2024 at 10:01 PM
4/6
We conducted experiments on forgetting random samples and entire features from the training set, consistently outperforming unlearning baselines and protecting user privacy.
We conducted experiments on forgetting random samples and entire features from the training set, consistently outperforming unlearning baselines and protecting user privacy.
3/6
Key Idea: We compare cached end-of-training gradients to those on the remaining data to identify parameters in the model to reset.
Key Idea: We compare cached end-of-training gradients to those on the remaining data to identify parameters in the model to reset.
December 14, 2024 at 10:00 PM
3/6
Key Idea: We compare cached end-of-training gradients to those on the remaining data to identify parameters in the model to reset.
Key Idea: We compare cached end-of-training gradients to those on the remaining data to identify parameters in the model to reset.
2/6
Key Motivation: In unlearning, we typically require access to the set of data being forgotten. How can we unlearn, without that data?
Key Motivation: In unlearning, we typically require access to the set of data being forgotten. How can we unlearn, without that data?
December 14, 2024 at 10:00 PM
2/6
Key Motivation: In unlearning, we typically require access to the set of data being forgotten. How can we unlearn, without that data?
Key Motivation: In unlearning, we typically require access to the set of data being forgotten. How can we unlearn, without that data?