Main content area

Applications of the Rolling Pin Method. 1. An Efficient Alternative to Bayesian Network Modeling and Inference

Mohseni Ahooyi, Taha, Arbogast, Jeffrey E., Soroush, Masoud
Industrial & Engineering Chemistry Research 2015 v.54 no.16 pp. 4316-4325
Bayesian theory, chemistry, data collection, engineering, probabilistic models, rolling
This paper presents a novel probabilistic modeling and inference method that is computationally more efficient than Bayesian networks (BNs). This method is applicable to systems with continuous variables. It is also applicable to systems with discrete variables that have an adequately high number of states. This work, indeed, represents an application of the rolling pin method introduced in our earlier paper [Mohseni Ahooyi et al., Ind. Eng. Chem. Res, DOI: 10.1021/ie503584q]. Unlike BNs, the method does not require knowledge of the causal relationships among the variables, because the rolling pin method models joint probabilities without causal factorization of the distributions. Furthermore, learning and inference steps are performed much faster by this method than by BNs, and the method enables one to perform local inference over any set of query variables; it allows for probabilities to be calculated over regions with no historical data, and it prevents information loss and an increase in the computational cost, both due to discretization of continuous variables. The application and performance of the method are shown through two examples.