Multi-task Transfer Learning for Bayesian Network Structures

We consider the interest of leveraging information between related tasks for learning Bayesian network structures. propose a new algorithm called Multi-Task Max-Min Hill Climbing (MT-MMHC) that combines ideas from transfer learning, multi-task constraint-based and search-and-score techniques. This approach consists in two main phases. The first one identifies most similar uses their similarity to learn corresponding undirected graphs. second directs edges with Greedy Search combined Branch-and-Bound algorithm. Empirical evaluation shows MT-MMHC can yield better results than structures individually or state-of-the-Art MT-GS terms structure accuracy computational time.