Learning Bayesian Networks in the Space of Structures by a Hybrid Optimization Algorithm

Abstract

Bayesian networks (BNs) are one of the most widely used class for machine learning and decision making tasks especially in uncertain domains. However, learning BN structure from data is a typical NP-hard problem. In this paper, we present a novel hybrid algorithm for BN structure learning, called MMABC. It's based on a recently introduced meta-heuristic, which has been successfully applied to solve a variety of optimization problems: Artificial Bee Colony (ABC). MMABC algorithm consists of three phases: (i) obtain an initial undirected graph by the subroutine MMPC. (ii) Generate the initial population of solutions based on the undirected graph and (iii) perform the ABC algorithm to orient the edges. We describe all the elements necessary to tackle our learning problem, and experimentally compare the performance of our algorithm with two state-of-the-art algorithms reported in the literature. Computational results demonstrate that our algorithm achieves better performance than other two related algorithms.

Bayesian network quantization method and structural learning

“…In Yang et al (2016), the bacterial foraging optimization algorithm is used to perform search and score-based structural learning of BNs. Finally, in Zhu, Liu and Jiang (2016), the artificial bee colony algorithm is used to perform search and score-based structural learning of BNs.…”