TabularQGAN: A quantum generative model for tabular data synthesis
| Source: arXiv AI
Tags: quantum computing, GAN, synthetic data, MIMIC-III, tabular data
A quantum GAN variant (TabularQGAN) achieves competitive performance against CTGAN and VAE-GMM on healthcare tabular data synthesis, filling a gap in quantum generative models that previously only handled homogeneous data — though results are limited to noiseless classical simulators, not real quantum hardware.
Details
Most quantum generative models handle homogeneous data; TabularQGAN extends the GAN architecture to mixed tabular data typical of enterprise and healthcare settings. Using a novel quantum circuit ansatz with flexible data encoding, the model was tested on the MIMIC-III clinical dataset and Adult Census benchmark against four classical baselines: CTGAN, CopulaGAN, VAE-GMM, and the LLM-based be-GReaT framework.\n\nOn MIMIC-III, TabularQGAN achieves competitive overall similarity scores per the SDMetrics library and outperforms classical baselines in some cases. The paper also introduces two custom generalization metrics that evaluate the model's ability to produce novel and useful tabular samples beyond the training distribution.\n\nKey caveat: all experiments run on a noiseless statevector simulator on classical hardware using reduced feature subsets. Real quantum hardware would introduce noise, and the proof-of-concept has not addressed scaling to full feature sets. The paper was published in Scientific Reports 2026.