Adapting existing BKB structures using new data

Bayesian knowledge bases (BKB) are a rule based probabilistic model that extends the well-known Bayes networks (BN), by naturally allowing for context-specific independence and for cycles in the directed graph. The learning process of BKB structures from large datasets consumes enormous amount of computational resources, even when using the somewhat simplified minimum description length (MDL) scoring. When a BKB structure exists for a dataset, adapting the existing structures can be used to expedite the learning process of for datasets that are known to be derived from similar causal structure. Empirical results show that the adaptation method is capable of successfully learning BKB structures that accurately represent the new data, are simple, and retain much of the existing structures.