Bayesian Generative Adversarial Networks via Gaussian Approximation for Tabular Data Synthesis

2026-02-25

AI Summary

This work proposes GACTGAN, a novel Bayesian generative adversarial network for tabular data synthesis that effectively balances utility and privacy. Addressing the limitations of conventional Conditional Tabular GANs (CTGANs)—which struggle to preserve both data fidelity and privacy—and the high computational costs of existing Bayesian GAN approaches that are ill-suited for tabular structures, GACTGAN introduces Stochastic Weight Averaging-Gaussian (SWAG) into the CTGAN generator. This integration enables an efficient Gaussian approximation of the posterior weight distribution with substantially reduced post-training computational and storage overhead. By leveraging SWAG, GACTGAN maintains the original data’s structural and statistical properties while enhancing synthetic data quality and mitigating privacy leakage risks, thereby achieving high-fidelity, efficient, and privacy-preserving tabular data generation.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Natural Language Processing: Generation
• Machine Learning: Privacy

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Web Mining and Content Analysis: Web data generation and simulation
• Security and Privacy: Data transparency and provenance

Abstract

Generative Adversarial Networks (GAN) have been used in many studies to synthesise mixed tabular data. Conditional tabular GAN (CTGAN) have been the most popular variant but struggle to effectively navigate the risk-utility trade-off. Bayesian GAN have received less attention for tabular data, but have been explored with unstructured data such as images and text. The most used technique employed in Bayesian GAN is Markov Chain Monte Carlo (MCMC), but it is computationally intensive, particularly in terms of weight storage. In this paper, we introduce Gaussian Approximation of CTGAN (GACTGAN), an integration of the Bayesian posterior approximation technique using Stochastic Weight Averaging-Gaussian (SWAG) within the CTGAN generator to synthesise tabular data, reducing computational overhead after the training phase. We demonstrate that GACTGAN yields better synthetic data compared to CTGAN, achieving better preservation of tabular structure and inferential statistics with less privacy risk. These results highlight GACTGAN as a simpler, effective implementation of Bayesian tabular synthesis.

Problem

• Tabular Data Synthesis
• Bayesian GAN
• Risk-Utility Trade-off
• Computational Overhead
• Privacy Risk

Innovation

• Bayesian GAN
• Gaussian Approximation
• Stochastic Weight Averaging-Gaussian
• Tabular Data Synthesis
• CTGAN