Summary of Incentives in Private Collaborative Machine Learning, by Rachael Hwee Ling Sim et al.
Incentives in Private Collaborative Machine Learning
by Rachael Hwee Ling Sim, Yehong Zhang, Trong Nghia Hoang, Xinyi Xu, Bryan Kian Hsiang Low, Patrick Jaillet
First submitted to arxiv on: 2 Apr 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 introduces a novel approach to collaborative machine learning that incentivizes participation from multiple parties while preserving privacy. The existing data valuation methods are extended to incorporate differential privacy (DP) as an incentive, allowing each party to select its required DP guarantee and perturb its sufficient statistic accordingly. The mediator values the perturbed statistic by the Bayesian surprise it elicits about the model parameters, enforcing a trade-off between privacy and valuation. This approach ensures that parties are deterred from selecting excessive DP guarantees that reduce the utility of the grand coalition’s model. Empirical results demonstrate the effectiveness and practicality of this approach on synthetic and real-world datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps make it safe for multiple groups to work together and share their data. The current way of doing this doesn’t consider how much privacy each group has, so they might not want to share as much as they could. To fix this, the researchers added something called differential privacy (DP) that makes sure each group’s data is kept private enough. They also came up with a new way to value and reward each group based on how surprised they are by what the model learned. This approach makes it so groups don’t want to sacrifice too much privacy for their own benefit. The results show that this works well in real-world scenarios. |
Keywords
* Artificial intelligence * Machine learning
