Ancestral Causal Inference

By Sara Magliacane, Tom Claassen, Joris M. Mooij. 2016

70 citations recorded by the source corpus.

Constraint-based causal discovery from limited data is a notoriously difficult challenge due to the many borderline independence test decisions. Several approaches to improve the reliability of the predictions by exploiting redundancy in the independence information have been proposed recently. Though promising, existing approaches can still be greatly improved in terms of accuracy and scalability. We present a novel method that reduces the combinatorial explosion of the search space by using a more coarse-grained representation of causal information, drastically reducing computation time. Additionally, we propose a method to score causal predictions based on their confidence. Crucially, our implementation also allows one to easily combine observational and interventional data and to incorporate various types of available background knowledge. We prove soundness and asymptotic consistency of our method and demonstrate that it can outperform the state-of-the-art on synthetic data, achieving a speedup of several orders of magnitude. We illustrate its practical feasibility by applying it on a challenging protein data set.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Domain Adaptation by Using Causal Inference to Predict Invariant Conditional Distributions (2017, methodology): “We use the method proposed by Magliacane et al. [2016] to query for the conﬁdence of whether some statement (e.g., Y ⊥⊥ C 1 | A ) is true or false.”
Joint Causal Inference from Multiple Contexts (2016, methodology): “For other methods (e.g., FCI (Spirtes et al., 1999; Zhang, 2008a) or methods that focus on ancestral relations, e.g., ACI (Magliacane et al., 2016b)), incorporating all background knowledge is less straightforward and as far as we know cannot be done with off-the-shelf implementations.”
Discovery of Causal Models that Contain Latent Variables through Bayesian Scoring of Independence Constraints (2017, methodology): “However, these hybrid methods, except [8,24,26,34], do not include the possibility that the CBNs being modeled contain hidden variables.”
Causal Transfer Learning (2017, methodology): “We use the method proposed in Magliacane et al. [2016] to query for the conﬁdence of whether some statement (e.g., Y ⊥⊥ I 1 | A ) is true or false.”
Permutation-Based Causal Structure Learning with Unknown Intervention Targets (2019, background): “, when there are no latent confounders, and ACI (Magliacane et al., 2016), HEJ (Hyttinen et al.”
Constraint-based Causal Discovery for Non-Linear Structural Causal Models with Cycles and Latent Confounders (2018, unknown): “In our experiments, we will use the weights proposed in [21]: λj = log pj − logα, where pj is the p-value of a statistical test with independence as null hypothesis, and α is a significance level (e.”
Permutation-based Causal Inference Algorithms with Interventions (2017, methodology): “Methods such as ACI [13], HEJ [10], COmbINE [25] and ICP [15] allow for latent confounders with possibly unknown interventional targets.”