Summary of Label-agnostic Forgetting: a Supervision-free Unlearning in Deep Models, by Shaofei Shen et al.
Label-Agnostic Forgetting: A Supervision-Free Unlearning in Deep Models
by Shaofei Shen, Chenhao Zhang, Yawen Zhao, Alina Bialkowski, Weitong Tony Chen, Miao Xu
First submitted to arxiv on: 31 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 educators can consider this paper as an innovative approach to machine unlearning, which aims to remove information derived from forgotten data while preserving that of the remaining dataset. The proposed method, Label-Agnostic Forgetting (LAF), operates without requiring labels during the unlearning process, making it suitable for large-scale datasets where annotation costs can be impractical. LAF leverages a variational approach to approximate representation distributions and adapts the original model to eliminate information from forgotten data at the representation level. A contrastive loss is introduced to facilitate matching between remaining data representations and those of the original model, ensuring predictive performance preservation. Experimental results across various unlearning tasks demonstrate LAF’s effectiveness without relying on full supervision information, outperforming state-of-the-art methods in semi-supervised scenarios. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Machine unlearning aims to remove forgotten data while preserving remaining dataset information. A new approach called Label-Agnostic Forgetting (LAF) doesn’t need labels during the process. LAF is like a filter that removes unwanted information and keeps important parts. It works by guessing what forgotten data looks like, so it can be removed. The method also makes sure the model stays good at making predictions. Tests show LAF is effective and even better when only some supervision information is available. |
Keywords
» Artificial intelligence » Contrastive loss » Machine learning » Semi supervised
