Causal Razors and Causal Search Algorithms

When performing causal discovery, assumptions have to be made on how the true causal mechanism corresponds to the underlying joint probability distribution. These assumptions are labeled as causal razors in this work. More than a dozen of causal razors have been surveyed in the literature. Offering a comprehensive logical analysis of them is the first task of this dissertation. Next, I demonstrate how edge counting is different from the enumeration of parameters when defining causal razors. This distinction gives rise to not only a practical dilemma for algorithm users but also an epistemological problem on what to assume. Another project that will be covered in this work is the extensive exposition of a recent causal search algorithm named GRaSP which outperforms certain state-of-the-art algorithms in accuracy and scalability. The algorithm will be examined in relation to the causal razors to establish its theoretical foundation.