The Synthetic Data Vault (SDV) uses a recursive algorithm to create a generative model of a relational database. First, it models the relationships between tables, creating an "extended table" by aggregating statistics from child tables into parent tables. Then, it uses a multivariate Gaussian Copula to model the correlations between columns (both original and aggregated) within each extended table. Finally, the algorithm can synthesize new data by sampling from these learned models, starting with parent tables and recursively generating corresponding child table rows.[^1] [^1]: Patki, N., Wedge, R., & Veeramachaneni, K. (2016, October). The synthetic data vault. In 2016 IEEE international conference on data science and advanced analytics (DSAA) (pp. 399-410). IEEE.