Elements of Causal Inference: Foundations and Learning Algorithms

By Jan Peters, Dominik Janzing, Bernhard Schölkopf. 2017

2,296 citations recorded by the source corpus.

How researchers see this paper

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

In Causal Inference & Discovery

Citations in the corpus

A Sober Look at the Unsupervised Learning of Disentangled Representations and their Evaluation (2020, background): “One potential approach to include inductive biases, offer interpretability, and generalization is the concept of independent causal mechanisms and the framework of causal inference (Pearl, 2009; Peters et al., 2017).”
Disentanglement via Mechanism Sparsity Regularization: A New Principle for Nonlinear ICA (2021, methodology): “The learned CGM could subsequently be used to simulate interventions on semantic variables (Pearl, 2009; Peters et al., 2017), such as changing the torque of the robot or the weight of the ball.”
Differentiable Causal Discovery from Interventional Data (2020, background): “These purely observational data sets were generated from the same CGM as the other data set containing interventions and had the same total sample size.”
An Analysis of the Adaptation Speed of Causal Models (2020, background): “These families are interesting because the direction is not identifiable from observational data (Peters et al., 2017, Ch.4) – e.g. pθ→ and pθ← can model the same set of distributions – which makes them challenging for causal discovery.”
Gradient-Based Neural DAG Learning (2019, methodology): “In this work, we propose a novel score-based method [5, 12] for structure learning named GraN-DAG which makes use of a recent reformulation of the original combinatorial problem of finding an optimal DAG into a continuous constrained optimization problem.”
Toward Causal Representation Learning (2021, background): “This principle entails several notions important to causality, including separate intervenability of causal variables, modularity and autonomy of subsystems, and invariance [183], [188].”
What do we need to build explainable AI systems for the medical domain? (2017, background): “This is important, because for decision support it is necessary to understand the causality of learned representations (Pearl, 2009; Gershman et al., 2015; Peters et al., 2017).”
A Causal Bayesian Networks Viewpoint on Fairness (2018, neutral): “Before doing that, we give some background on CBNs [18,39,40,42,45], assuming that all variables except A are continuous.”
Path-Specific Counterfactual Fairness (2018, methodology): “…& Sh-pitser, 2018; Russell et al., 2017; Zhang et al., 2017; Zhang & Wu, 2017; Zhang & Bareinboim, 2018) that using a causal framework (Pearl, 2000; Dawid, 2007; Pearl et al., 2016; Peters et al., 2017) would lead to a more intuitive, powerful, and less error-prone way of reasoning about fairness.”
Towards Causal Representation Learning (2021, background): “situation is summarized in Table I, adapted from [188].”
On the Binding Problem in Artificial Neural Networks (2020, background): “This is especially relevant due to the connection between systematic generalization and the increased robustness when considering so-called independent causal mechanisms (Peters et al., 2017).”
Learning explanations that are hard to vary (2020, background): “A fundamental motivation for our work comes from causality (Pearl, 2009; Peters et al., 2017), based on the idea that statistical dependencies are epiphenomena of an underlying causal model.”
Improving the accuracy of medical diagnosis with causal machine learning (2020, methodology): “Note that this probability refers to two contradictory states of S and so cannot be represented as a standard posterior 49,59 .”
Algorithmic recourse under imperfect causal knowledge: a probabilistic approach (2020, background): “While for many applications it is possible to draw a causal diagram from expert knowledge, assumptions about the form of structural equations are, in general, not testable and may thus not hold in practice [38].”
Simpson's Paradox in COVID-19 Case Fatality Rates: A Mediation Analysis of Age-Related Causal Effects (2020, background): “Mediation formulas: effects in causally sufficient systems For causally sufficient systems, the interventional distributions of each variable given its causal parents equal the corresponding observational distributions [16], corresponding to the intuition that they represent mechanisms rather than m”
A theory of independent mechanisms for extrapolation in generative models (2020, methodology): “However, once such network has been trained on a particular dataset, can it can be leveraged to simulate distributions with meaningful differences? A causal model allows the different mechanisms involved in generating the data to be intervened on independently, based on the principle of Independent ”
Weakly-Supervised Disentanglement Without Compromises (2020, background): “On the other hand, many data modalities are not observed as i.i.d. samples from a distribution (Dayan, 1993; Storck et al., 1995; Hochreiter & Schmidhuber, 1999; Bengio et al., 2013; Peters et al., 2017; Thomas et al., 2017; Schölkopf, 2019).”
Causal Models for Dynamical Systems (2020, background): “There is a close connection between experimentation and causal inference [Imai et al., 2013, Peters et al., 2017]; the selection of measurement readouts, time points or intervention strategies could guide experimentation and has the potential to signiﬁcantly reduce the number of complicated and…”
Causality for Machine Learning (2019, background): “My understanding of causality has been shaped by Judea Pearl and a number of collaborators and colleagues, and much of it went into a book written with Dominik Janzing and Jonas Peters (Peters et al., 2017).”
Learning Neural Causal Models from Unknown Interventions (2019, background): “A Structural Causal Model (SCM) (Peters et al., 2017) over a ﬁnite number M of random variables X i is a set of structural assignments where N i is jointly-independent noise and pa ( i, C ) is the set of parents (direct causes) of variable i under conﬁguration C of the SCM directed acyclic graph,…”