Abstract
We consider the problem of learning structured causal models from observational data. In this work, we use causal Bayesian networks to represent causal relationships among model variables. To this effect, we explore the use of two types of independencies—context-specific independence (CSI) and mutual independence (MI). We use CSI to identify the candidate set of causal relationships and then use MI to quantify their strengths and construct a causal model. We validate the learned models on benchmark networks and demonstrate the effectiveness when compared to some of the state-of-the-art Causal Bayesian Network Learning algorithms from observational Data.
Highlights
Given the recent success of machine learning, deep learning, in several applications (Goodfellow et al, 2016), there is an increased interest in learning more explainable models including causal models.Many researchers have attempted to develop methods to infer causality from observational data over for several years (Pearl, 1988b, 2000; Neapolitan et al, 2004)
In order to develop an approach for this motivating application, we propose an algorithm for learning a Bayesian network (BN) from dependency networks (DN), that can scale to a large number of variables
These are the learned networks obtained by our approach DN2CN and baseline methods PC, Fast Greedy Equivalence Search (FGES) & Fast Causal Inference (FCI) summarized together with the ground truth network
Summary
Given the recent success of machine learning, deep learning, in several applications (Goodfellow et al, 2016), there is an increased interest in learning more explainable models including causal models.Many researchers have attempted to develop methods to infer causality from observational data over for several years (Pearl, 1988b, 2000; Neapolitan et al, 2004). Recent advances in the field of discovering causality has looked at learning Causal Bayesian Network (CBN). In this framework, causations among variables are represented with a Directed Acyclic Graph (DAG) (Pearl, 2000). The problem of learning a DAG from data is not computationally realistic as the number of possible DAGs grows exponentially with the number of nodes. This computational complexity has prevented the adaptation and application of causal discovery approaches to high dimensional datasets, with a few examples. Hybrid learning approaches combine the advantages of both approaches; for example, using constraint-based techniques to estimate the network skeleton, and using score-based techniques to identify the set of edge orientations that best fit the data (Tsamardinos et al, 2006)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.