Summary of Differentially Private Neural Network Training Under Hidden State Assumption, by Ding Chen et al.
Differentially Private Neural Network Training under Hidden State Assumption
by Ding Chen, Chen Liu
First submitted to arxiv on: 11 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 This paper proposes a new method called differentially private stochastic block coordinate descent (DP-SBCD) for training neural networks while ensuring differential privacy under the hidden state assumption. The approach combines Lipschitz neural networks with decomposed training of specific layers, extending existing analyses to non-convex problems and proximal gradient descent algorithms. A novel aspect is the use of calibrated noise from adaptive distributions, achieving improved empirical trade-offs between utility and privacy. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps us train artificial intelligence models in a way that keeps people’s private information safe. It uses a new technique called DP-SBCD to make sure neural networks are trained privately. The method breaks down the training process into smaller parts, like training individual layers, which makes it easier to analyze how private the model is. This approach also improves the balance between keeping data private and making the model work well. |
Keywords
* Artificial intelligence * Gradient descent
