Causal Models as Minimal Descriptions of Multivariate Systems

By Jan Lemeire, E. Dirkx. 2006

Subfields: Probabilistic & Bayesian Methods, Causal Inference & Discovery

How researchers see this paper

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

In Probabilistic & Bayesian Methods

In Causal Inference & Discovery

Citations in the corpus

Discovering Causal Signals in Images (2016, background): “In this example, ICM is satisfied under the correct causal direction X → Y , but violated under the wrong causal direction Y → X [3].”
Semi-supervised learning, causality, and the conditional cluster assumption (2019, background): “For the bivariate setting, ICM reduces to an independence of cause and mechanism [4, 22].”
On causal and anticausal learning (2012, methodology): “2 This assumption has been used by Janzing & Sch¨olkopf (2010), inspired by Lemeire & Dirkx (2007).”
Identifiability of Causal Graphs using Functional Models (2011, background): “Consider the DAG G̃′ that equals G′ with additional edges Y → W and W ′ → W for all Y ∈ Y and W ′ ∈W \ {W}.”
Inferring deterministic causal relations (2010, background): “While this postulate is informal in the sense that we do not specify what we mean by independence, we point out that Lemeire and Dirkx [2006] and Janzing and Sch¨olkopf [2008] give a formalization in terms of algorithmic independence.”
Causal Inference Using the Algorithmic Markov Condition (2008, methodology): “As already noted by [37], the independence of mechanisms is related to Pearl’s thoughts on the stability of causal sta tements: the causal mechanism P (Xj |PAj) does not change if one changes the input distribution P (PAj) by influencing the variablesPAj .”
Telling cause from effect based on high-dimensional observations (2009, background): “The underlying idea is that the causal hypothesis X → Y is only acceptable if the shortest description of the joint distribution P (X,Y ) is given by separate descriptions of the input distribution P (X) and the conditional distribution P (Y |X) [10], expressing the fact that they represent independ”
Testing whether linear equations are causal: A free probability theory approach (2011, background): “Janzing et al. (2010) shows that pairs (A,ΣX) satisfying τn(A TAΣX) ≈ τn(ATA)τn(ΣX) (4) 1For random variables, Lemeire and Dirkx (2006) and Janzing and Schölkopf (2010) postulate that the shortest description of P (cause, effect) is given by separate descriptions of P (cause) and P (effect|cause).”
Distinguishing Cause and Effect via Second Order Exponential Models (2009, methodology): “[15] proposed to prefer those DAGs as causal hypotheses for which the shortest description of the joint density p ( x 1 , . . . , x n ) is given by separate descriptions of causal conditionals p ( x j | pa j ) in eq.”
Causal Regularization (2017, background): “…between the cause and eﬀect based on the natural assumption that steps in the process that generates the data are independent from each other, see (Lemeire and Dirkx, 2006; Janzing et al., 2012; Daniusis et al., 2010; Lopez-Paz, 2016; Chalupka et al., 2016; Kocaoglu et al., 2016) and the…”
Probabilistic latent variable models for distinguishing between cause and effect (2010, methodology): “If complexity is measured in terms of Kolmogorov complexity, this kind of reasoning would be in the spirit of the principle of “algorithmically independent conditionals” [11], which can also be embedded into a general theory of algorithmic-information-based causal discovery [12].”
Justifying Additive Noise Model-Based Causal Discovery via Algorithmic Information Theory (2009, methodology): “…based causal discovery thus is an application of two recent proposals for using algorithmic information theory in causal inference: [9] postulated that the shortest description of P (cause , eﬀect) is given by separate descriptions of P (cause) and P (eﬀect | cause), which would be…”
From Dependence to Causation (2016, background): “In the spirit of Occam’s razor, we prefer the conditional decomposition that provides with the shortest description of the causal structure contained in the joint distribution p(x,y) (Lemeire and Dirkx, 2006).”
Information-geometric approach to inferring causal directions (2012, background): “Extending an idea in [4,5] postulates that X → Y is only acceptable as causal hypothesis if the shortest description of P X,Y is given by separate descriptions of P Y |X and P X .”
Causal Inference via Kernel Deviance Measures (2018, methodology): “First, we take an information-theoretic approach to reasoning about the complexity of distributions similar to [11, 10].”
On the Entropy Production of Time Series with Unidirectional Linearity (2009, result): “Similar arguments have been used in causal inference recently [19, 20].”
Telling Cause from Effect Using MDL-Based Local and Global Regression (2017, background): “in terms of Kolmogorov complexity, K(P (X,Y )), will be shorter when first describing the distribution of the cause K(P (X)) and than describing the distribution of the effect given the cause K(P (Y | X)) than vice versa [9], [12].”
Identiﬁability of Causal Graphs using Functional Models (background): “…it can only happen if the cause distribution p (cause) and the conditional distribution of the eﬀect given the cause p (eﬀect | cause) are matched in a precise way, whereas one rather expects input and mechanism to be most often “independent” [Lemeire and Dirkx, 2006, Janzing and Sch¨olkopf, 2010].”
Robust Learning via Cause-Effect Models (2011, background): “We are interested in the question how robust (or invariant) the estimation is with respect to changes in the noise variables of the underlying functional causal model.”