Summary of Partially Blinded Unlearning: Class Unlearning For Deep Networks a Bayesian Perspective, by Subhodip Panda and Shashwat Sourav and Prathosh A.p
Partially Blinded Unlearning: Class Unlearning for Deep Networks a Bayesian Perspective
by Subhodip Panda, Shashwat Sourav, Prathosh A.P
First submitted to arxiv on: 24 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computer Vision and Pattern Recognition (cs.CV)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Machine learning models must eliminate information derived from specific subsets of user training data that can no longer be utilized to comply with regulatory standards. This led to the emergence of Machine Unlearning as a crucial area of research. The goal is to selectively discard information linked to specific data classes from pre-trained models, eliminating the need for extensive retraining. We framed the class unlearning problem from a Bayesian perspective, yielding a loss function that minimizes log-likelihood associated with unlearned data and incorporates stability regularization. Our novel approach, Partially-Blinded Unlearning (PBU), outperforms existing methods without requiring awareness of the entire training dataset. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Machine learning models need to forget specific information from user data to follow regulations. This requires a new area of research called Machine Unlearning. The goal is to make models forget specific things about certain types of data, so we don’t have to retrain them from scratch. We used a special way of looking at the problem, which gave us a formula that helps the model forget unwanted information. Our new approach works better than others without needing to see all the training data. |
Keywords
* Artificial intelligence * Log likelihood * Loss function * Machine learning * Regularization
