Generating Benchmark Health Data Using a Tabular Diffusion Transformer
2026-08-14 • Machine Learning
Machine LearningArtificial Intelligence
AI summaryⓘ
The authors created a method to generate new synthetic datasets from several different tables that have different kinds of information. First, they turn each original table into a common format summarizing key statistics like distributions and correlations. Then, they train a special model to learn patterns from these summary tables and generate new statistical tables. Finally, they transform these summaries back into new synthetic raw data. Their experiments show this method produces realistic and varied synthetic data, even when starting from diverse tables.
cross-tabular data generationheterogeneous tablesmarginal distributionpairwise correlationdiffusion transformersynthetic data generationmultivariate Gaussian samplingprobability integral transformstatistical tables
Authors
Hao Yan, Lisa Pilgram, Dan Liu, Linglong Kong, Fida Dankar, Khaled El Emam
Abstract
Cross-Tabular Data Generation (CTDG) seeks to learn a generative model from multiple heterogeneous tables and produce new synthetic tabular datasets. However, existing synthetic tabular data generation methods are largely restricted to single-input-table scenarios and struggle to effectively handle multiple heterogeneous tables with diverse feature sets. To address this limitation, we propose a two-stage framework for cross-tabular data generation. In the first stage, each heterogeneous raw table is transformed into a standardized statistical table with the same set of columns across all tables. Each statistical table captures the marginal distributions of the original columns and the pairwise correlations among them. In the second stage, a diffusion transformer model is trained to capture structural patterns across these homogeneous statistical tables and to generate synthetic statistical tables. Synthetic raw tables are subsequently reconstructed from the generated statistical tables via multivariate Gaussian sampling followed by an inverse probability integral transform. This two-stage CTDG framework enables the learning of a unified generative model from multiple heterogeneous tables and supports the generation of an unlimited number of realistic synthetic heterogeneous tables. Experimental results demonstrate high fidelity in the learned statistical representations and a favorable fidelity-diversity trade-off in the generated synthetic data, validating the effectiveness of the proposed approach.