K2GA: Heuristically Guided Evolution of Bayesian Network Structures from Data

— We present K2GA, an algorithm for learning Bayesian network structures from data. K2GA uses a genetic algorithm to perform stochastic search, while employing a modified version of the K2 heuristic to score proposed networks and improve future generations. We show each component of K2GA, a combination of these components to form the basic algorithm, extensions to the algorithm for improved accuracy, and numerical results.

• Type: Conference
• Year: 2007
• Where: CIDM
• Authors: Eli Faulkner
• Keywords: Algorithm, Artificial Intelligence, Bayesian Network Structures, CIDM 2007, Genetic Algorithm