Summary of Balls-and-bins Sampling For Dp-sgd, by Lynn Chua et al.
Balls-and-Bins Sampling for DP-SGD
by Lynn Chua, Badih Ghazi, Charlie Harrison, Ethan Leeman, Pritish Kamath, Ravi Kumar, Pasin Manurangsi, Amer Sinha, Chiyuan Zhang
First submitted to arxiv on: 21 Dec 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Cryptography and Security (cs.CR); Data Structures and Algorithms (cs.DS); Machine Learning (stat.ML)
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 introduces a new method for differentially private (DP) optimization called Balls-and-Bins sampling, which is used in DP-SGD implementations. This approach combines the benefits of shuffling and Poisson subsampling, achieving both high utility and privacy amplification. The authors demonstrate that models trained using this method have similar performance to those trained with traditional shuffling or Poisson subsampling at the same noise multiplier. However, Balls-and-Bins sampling enjoys better privacy amplification in practical regimes. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about a new way to make sure computer programs are private and safe when they’re training models. Right now, there are two common ways to do this: shuffling and Poisson subsampling. But researchers found that one of these methods can actually be more private than the other in some cases. So, the authors came up with a new method called Balls-and-Bins sampling that combines the good parts of both. This new method works well and is just as good at protecting privacy as the other two, but it’s better in certain situations. |
Keywords
» Artificial intelligence » Optimization
