Discovering causal relationships from observational data is a crucial problem and has applications in many research areas. PC algorithm is the state-of-the-art method in the constraint based approach. However, the PC algorithm is worst-case exponential to the number of nodes (variables), and thus it is inefficient when applying to high dimensional data, e.g. gene expression datasets where the causal relationships between thousands of nodes (genes) are explored. In this paper, we propose a fast and memory efficient PC algorithm using the parallel computing technique. We apply our method on a range of synthetic and real-world high dimensional datasets. Experimental results on a dataset from DREAM 5 challenge show that the PC algorithm could not produce any results after running more than 24 hours; meanwhile, our parallel-PC algorithm with a 4-core CPU computer managed to finish within around 12.5 hours, and less than 6 hours with a 8-core CPU computer.
View on arXiv