TY - CONF
T1 - On the Complexity of Identification in Linear Structural Causal Models
AU - Dörfler, Julian
AU - van der Zander, Benito
AU - Bläser, Markus
AU - Liśkiewicz, Maciej
N1 - Publisher Copyright:
© 2024 Neural information processing systems foundation. All rights reserved.
PY - 2024
Y1 - 2024
N2 - Learning the unknown causal parameters of a linear structural causal model is a fundamental task in causal analysis. The task, known as the problem of identification, asks to estimate the parameters of the model from a combination of assumptions on the graphical structure of the model and observational data, represented as a non-causal covariance matrix. In this paper, we give a new sound and complete algorithm for generic identification which runs in polynomial space. By a standard simulation result, namely PSPACE ⊆ EXP, this algorithm has exponential running time which vastly improves the state-of-the-art double exponential time method using a Gröbner basis approach. The paper also presents evidence that parameter identification is computationally hard in general. In particular, we prove, that the task asking whether, for a given feasible correlation matrix, there are exactly one or two or more parameter sets explaining the observed matrix, is hard for ∀R, the co-class of the existential theory of the reals. In particular, this problem is coNP-hard. To our best knowledge, this is the first hardness result for some notion of identifiability.
AB - Learning the unknown causal parameters of a linear structural causal model is a fundamental task in causal analysis. The task, known as the problem of identification, asks to estimate the parameters of the model from a combination of assumptions on the graphical structure of the model and observational data, represented as a non-causal covariance matrix. In this paper, we give a new sound and complete algorithm for generic identification which runs in polynomial space. By a standard simulation result, namely PSPACE ⊆ EXP, this algorithm has exponential running time which vastly improves the state-of-the-art double exponential time method using a Gröbner basis approach. The paper also presents evidence that parameter identification is computationally hard in general. In particular, we prove, that the task asking whether, for a given feasible correlation matrix, there are exactly one or two or more parameter sets explaining the observed matrix, is hard for ∀R, the co-class of the existential theory of the reals. In particular, this problem is coNP-hard. To our best knowledge, this is the first hardness result for some notion of identifiability.
M3 - Conference Papers
ER -