Summary of Towards Efficient Target-level Machine Unlearning Based on Essential Graph, by Heng Xu et al.
Towards Efficient Target-Level Machine Unlearning Based on Essential Graph
by Heng Xu, Tianqing Zhu, Lefeng Zhang, Wanlei Zhou, Wei Zhao
First submitted to arxiv on: 16 Jun 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Cryptography and Security (cs.CR)
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 Medium Difficulty Summary: Machine unlearning, an emerging technology, has gained attention due to regulatory concerns, privacy issues, and usability factors. Existing studies focus on forgetting instances or classes, but these approaches don’t scale for partial targets within instances. For instance, one might want to forget a person in all instances containing that person and other targets. Directly migrating instance-level unlearning to target-level unlearning can reduce model performance or fail to erase information completely. To address this, we propose “target unlearning”, which focuses on removing partial targets from the model. We construct an essential graph data structure using the model explanation method to describe relationships between important parameters. Then, we filter parameters important for remaining targets and use pruning-based unlearning to remove information about the target. Experiments with different models on various datasets demonstrate the effectiveness of our approach. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Low Difficulty Summary: Imagine you’re trying to forget something a computer learned. This is called “machine unlearning”. Right now, there’s a big need for this because of privacy concerns and laws that regulate how computers use information. Most studies focus on forgetting whole groups or classes, but this doesn’t work well when you want to forget just part of the information. For example, imagine wanting to forget a person in all pictures where they’re with other people. Currently, trying to do this can make the computer’s performance worse or not completely erase the information. Our solution is called “target unlearning” and it focuses on forgetting specific parts of the information. We use special graphs to keep track of important details and then remove the unwanted information. We tested our approach with different computers and datasets, and it worked well. |
Keywords
* Artificial intelligence * Attention * Pruning
