How similar is synthetic tabular data to the real thing?

This new approach uses KL and JS divergences to evaluate full joint distributions — not just feature by feature. A more reliable way to validate synthetic datasets in practice.

👉 synthema.eu/publications/

#SyntheticData #DataValidation #AI

Over 450 haematological diseases affect blood cells, the lymphatic system & coagulation, posing major challenges to public #HealthSystems.
#SYNTHEMA is developing a secure platform for synthetic #ClinicalData to enable GDPR-compliant research in rare #HaematologicalDiseases.

👉 synthema.eu

Can synthetic tabular data be more accurate?

This new model by researchers at Universidad Politécnica de Madrid combines a Variational Autoencoder with a Gaussian Mixture Model. Results outperform #CTGAN and #TVAE, especially on medical datasets.

👉 synthema.eu/publications/

#SYNTHEMA

Two productive days in San Sebastian for the #SYNTHEMA consortium have come to an end. Lively discussions, clear goals for the final year: a recap publication on synthetic data in haematology, joint training with @erneurobloodnet.bsky.social, new papers, and the SYNTHEMA platform launch

synthema.eu

Most synthetic data models assume data abundance. This study shows how transfer learning and meta-learning improve synthetic data generation for cancer research in low-data scenarios, where real clinical constraints apply.

👉 synthema.eu/publications/

#SYNTHEMA #SyntheticData #AI

Most synthetic data models assume data abundance. This study shows how transfer learning and meta-learning improve synthetic data generation for cancer research in low-data scenarios, where real clinical constraints apply.

👉 synthema.eu/publications/

#SYNTHEMA #CancerResearch

In #SYNTHEMA, data collection is central to building GDPR-compliant, AI-ready datasets for Sickle Cell Disease and Acute Myeloid Leukaemia. Harmonised across countries, the data supports synthetic generation, federated learning, and future rare disease research.

👉 synthema.eu/2025/09/23/d...