Welcome to the Functor Bayes Net system!
A newer version of our Bayes net learning algorithm is available here. The new version does not yet output Markov Logic Networks.
Methods for learning directed graphical models, or Bayes nets, from data have been the focus of decades of work in AI, statistics, philosophy, and social science. However, the majority of this work has focused on propositional data and do not address dependencies between entities and relationships in relational data.
We provide a software package for learning Bayes nets for relational data. The Bayes Net code basis was the open-source The TETRAD Project. TETRAD permits users to generate directed graphical statistical/causal models for non relational data. We extend the Tetrad search algorithms to allow learning in the relational setting.
A relational Bayes net can be converted to a Markov Logic Network in a standard way (moralization). Markov Logic networks are a prominent statistical-relational formalism invented by Pedro Domingos and his collaborators. The moralization approach is one of the fastest and most effective structure learning methods for Markov Logic Networks.
Contact: Oliver Schulte (oschulte AT cs.sfu.ca).