This academic article focuses on the challenge of preserving privacy while enabling collaborative data sharing. It discusses synthetic data generation as a solution, producing artificial data that mirrors the statistical properties of private data. The article introduces the study of federated synthetic tabular data generation, building upon a method known as AIM. The researchers present two new methods, DistAIM and FLAIM, demonstrating how to distribute AIM and showing that this can improve utility while reducing overhead. The article emphasizes the importance of mitigating privacy issues and maintaining a private proxy of heterogeneity.
Publication date: 6 Oct 2023
Project Page: Not provided
Paper: https://arxiv.org/pdf/2310.03447
