A Review on Approaches for Causal Structure Identification
Learning the skill to discover causal relations and to make use of them is said to be an essential step in human intelligence and potentially also in machine intelligence. The domain of causal discovery tackles the challenge of identifying causal structures from data collected from observations or experiments by exploiting special properties of causal relations. While current causality literature focuses on methods of probabilistic discovery using conditional independence tests and hard and soft interventions other lesser known approaches are neglected. In this work, we will give a short review on approaches for gaining causal knowledge and provide a categorization of methods. Also, we will introduce the Joint Discovery Assumption that is essential for combining different approaches for causal discovery. Finally, we discuss the open research fields we deduce from our categorization.