Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

A Recursive Markov Boundary-Based Approach to Causal Structure Learning

About

Constraint-based methods are one of the main approaches for causal structure learning that are particularly valued as they are asymptotically guaranteed to find a structure that is Markov equivalent to the causal graph of the system. On the other hand, they may require an exponentially large number of conditional independence (CI) tests in the number of variables of the system. In this paper, we propose a novel recursive constraint-based method for causal structure learning that significantly reduces the required number of CI tests compared to the existing literature. The idea of the proposed approach is to use Markov boundary information to identify a specific variable that can be removed from the set of variables without affecting the statistical dependencies among the other variables. Having identified such a variable, we discover its neighborhood, remove that variable from the set of variables, and recursively learn the causal structure over the remaining variables. We further provide a lower bound on the number of CI tests required by any constraint-based method. Comparing this lower bound to our achievable bound demonstrates the efficiency of the proposed approach. Our experimental results show that the proposed algorithm outperforms state-of-the-art both on synthetic and real-world structures.

Ehsan Mokhtarian, Sina Akbari, AmirEmad Ghassami, Negar Kiyavash• 2020

Related benchmarks

TaskDatasetResultRank
Causal Structure LearningSynthetic nD=10000, d=2, dmax=10, 100 nodes
CI Test Count5.48
13
Causal Structure LearningSynthetic nD=10000, d=2, dmax=10, 200 nodes
Number of CI Tests21.11
13
Causal Structure LearningSynthetic nD=10000, d=2, dmax=10, 400 nodes
CI Test Count82.07
13
Causal DiscoveryBinary data 10 nodes, nD=1000, d=2, dmax=10
Number of CI tests85.41
7
Causal DiscoveryBinary data 15 nodes, nD=1000, d=2, dmax=10
Number of CI Tests196.5
7
Causal DiscoveryBinary data 20 nodes, nD=1000, d=2, dmax=10
Number of CI Tests322.1
7
Local Causal DiscoveryLinear Gaussian 100 nodes
CI Test Count (x10^3)5.65e+3
7
Local Causal DiscoveryLinear Gaussian 200 nodes
CI Test Count (x10^3)21.99
7
Local Causal DiscoveryLinear Gaussian 400 nodes
Number of CI tests (x10^3)87.64
7
Optimal adjustment set identificationMAGIC-NIAB
CI Tests9.66e+3
6
Showing 10 of 11 rows

Other info

Follow for update