EQ is an unsupervised structural learning algorithm that explores the space of Equivalence Classes of Bayesian networks. This is a space of Partially Directed Acyclic Graphs (PDAGs), smaller that the space of Bayesian networks (DAGs), where the graphs consist of edges (i.e. undirected connections) and arcs. Edges are used when the direction of the relationship does not impact the represented joint probability distribution. The exploration of this space is carried out using a Greedy Search algorithm, i.e. the graph transformation chosen is the one that most decreases the MDL score.
New Feature: TabooEQ
This new unsupervised structural learning algorithm is an EQ algorithm where the exploration of the space is carried out using a Taboo Search algorithm. Instead of choosing the transformation that most decreases the score, the algorithm selects the best one, even if this transformation increases the score.