Summary of Machine Unlearning Using a Multi-gan Based Model, by Amartya Hatua et al.
Machine Unlearning using a Multi-GAN based Model
by Amartya Hatua, Trung T. Nguyen, Andrew H. Sung
First submitted to arxiv on: 26 Jul 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 The paper proposes a new machine unlearning approach that leverages multiple Generative Adversarial Network (GAN) based models. The method involves two phases: data reorganization using GAN-generated synthetic data with inverted class labels of the forget datasets, followed by fine-tuning a pre-trained model. The GAN models comprise two pairs of generators and discriminators that generate synthetic data for retain and forget datasets. A pre-trained model is used to obtain class labels for synthetic and original forget datasets, which are then inverted. The combined datasets are utilized to fine-tune the pre-trained model to achieve an unlearned model. Experiments were conducted on the CIFAR-10 dataset, testing the unlearned models using Membership Inference Attacks (MIA). The approach outperforms state-of-the-art models and standard unlearning classifiers. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper develops a new way to “unlearn” a machine learning model so it forgets what it has learned. To do this, they use a combination of Generative Adversarial Networks (GAN) and fine-tuning a pre-trained model. They create synthetic data that looks like the real data but is different enough to help the model learn new things. This helps the model perform better than other unlearning approaches. |
Keywords
» Artificial intelligence » Fine tuning » Gan » Generative adversarial network » Inference » Machine learning » Synthetic data
