Causal Effect Inference with Deep Latent-Variable Models
By Christos Louizos, Uri Shalit, Joris M. Mooij, David A. Sontag, Richard S. Zemel, Max Welling. 2017
908 citations recorded by the source corpus.
Learning individual-level causal effects from observational data, such as inferring the most effective medication for a specific patient, is a problem of growing importance for policy makers. The most important aspect of inferring causal effects from observational data is the handling of confounders, factors that affect both an intervention and its outcome. A carefully designed observational study attempts to measure all important confounders. However, even if one does not have direct access to all confounders, there may exist noisy and uncertain measurement of proxies for confounders. We build on recent advances in latent variable modeling to simultaneously estimate the unknown latent space summarizing the confounders and the causal effect. Our method is based on Variational Autoencoders (VAE) which follow the causal structure of inference with proxies. We show our method is significantly more robust than existing methods, and matches the state-of-the-art on previous benchmarks focused on individual treatment effects.
Subfields: Artificial Intelligence, Statistics and Probability, 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 Artificial Intelligence
- Sergey Levine
- Bernhard Schölkopf
- Yoshua Bengio
- Stefano Ermon
- Mohit Bansal
- Percy Liang
- Ruslan Salakhutdinov
- Pieter Abbeel
In Statistics and Probability
- Bernhard Schölkopf
- David M. Blei
- Mihaela van der Schaar
- Ioana Bica
- Stefano Ermon
- Ahmed M. Alaa
- Rajesh Ranganath
- Diederik P. Kingma
In Probabilistic & Bayesian Methods
- Stefano Ermon
- Jiaming Song
- David M. Blei
- Diederik P. Kingma
- Rajesh Ranganath
- Sergey Levine
- Max Welling
- Bernhard Schölkopf
In Causal Inference & Discovery
- Bernhard Schölkopf
- Dominik Janzing
- Michel Besserve
- Rémy Sun
- Christina Heinze-Deml
- Ioana Bica
- Jan Peters
- Julius von Kügelgen
Citations in the corpus
- Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Challenges toward Responsible AI (2019, neutral): “• Causality: Another common goal for explainability is that of finding causality among data variables.”
- Peeking Inside the Black-Box: A Survey on Explainable Artificial Intelligence (XAI) (2018, background): “as part of the loss function [77], or as part of the architecture structure, in terms of operations between layers [78], [79].”
- Explanation Methods in Deep Learning: Users, Values, Concerns and Challenges (2018, background): “as part of the loss function [12, 11], modules that add additional capabilities [49, 43], or as part of the architecture structure, in terms of operations between layers [32, 55, 33, 19].”
- Algorithmic recourse under imperfect causal knowledge: a probabilistic approach (2020, background): “An alternative approach could address causal sufficiency violations by relying on latent variable models to estimate confounders from multiple causes [61] or proxy variables [31], or to work with bounds on causal effects instead [4, 49].”
- A theory of independent mechanisms for extrapolation in generative models (2020, background): “Deep neural network have been leveraged in causal inference for learning causal graphs between observed variables (Lopez-Paz and Oquab 2016) and associated causal effects (Louizos et al. 2017; Shalit, Johansson, and Sontag 2017; Kocaoglu et al. 2017; Lachapelle et al. 2019; Zhu, Ng, and Chen 2019).”
- Time Series Deconfounder: Estimating Treatment Effects over Time in the Presence of Hidden Confounders (2019, methodology): “Alternative methods for dealing with hidden confounders in the static setting involve using proxy variables as noisy substitutes for latent confounders (Lash et al., 2014; Kuroki & Pearl, 2014; Louizos et al., 2017; Lee et al., 2018).”
- Deconfounding Reinforcement Learning in Observational Settings (2018, methodology): “However, using proxy variables to correctly recover causal effects requires strict mathematical assumptions (Louizos et al., 2017; Edwards et al., 2015; Kuroki and Pearl, 2014; Miao et al., 2018; Pearl, 2012; Wooldridge, 2009) and, in practice, we do not know whether or not the available data meets ”
- Conditional variance penalties and domain shift robustness (2017, methodology): “Louizos et al. (2017) propose the use of deep latent variable models and proxy variables to estimate individual treatment effects.”
- More Robust Doubly Robust Off-policy Evaluation (2018, background): “The OPE problem can also be viewed as a form of counterfactual reasoning to infer the causal effect of a new treatment from historical data (e.g., Bottou et al. 2013; Shalit et al. 2017; Louizos et al. 2017).”
- Distribution Matching in Variational Inference (2018, background): “, 2017), while posterior inference is used in a broad range of applications in representation learning in different domain applications such as reinforcement learning, fewshot learning and counter factual reasoning (Higgins et al., 2017b; Louizos et al., 2017; Mathieu et al., 2016; Rezende et al., 2”
- Learning Controllable Fair Representations (2018, methodology): “We evaluate our results on three datasets (Zemel et al., 2013; Louizos et al., 2017; Madras et al., 2018).”
- Algorithm fairness in artificial intelligence for medicine and healthcare (2023, background): “179), treatment-invariant representations for producing counterfactual explanations 259–261 , and race-invariant or ethnicity-invariant features to remove disparate impact in deep-learning models 67,101,262–268 .”
- Towards Realistic Individual Recourse and Actionable Explanations in Black-Box Decision Making Systems (2019, methodology): “The implementation provided here: https://github.com/AMLab-Amsterdam/ CEVAE is modified to incorporate conditioning for immutable variables.”
- Causal Discovery with Attention-Based Convolutional Neural Networks (2019, methodology): “Multiple deep learning models exist for non-temporal causal discovery: Variational Autoencoders [33] to estimate causal effects, Causal Generative Neural Networks to learn functional causal models [34] and the Structural Agnostic Model (SAM) [35] for causal graph reconstruction.”
- A Survey of Learning Causality with Data (2018, methodology): “s us to go beyondtheunconfoundednessassumption.WithspecificdeeplearningmodelssuchastheBalancing Counterfactual Regression [77], the TARnet [138], and the Causal Effect Variational Autoencoder (CEVAE) [97], we can learn representations z i of each instance i based on (x i,d i,y i). Here, we introduce”
- GENERATIVE ADVERSARIAL NETS (2018, methodology): “More recent works focus on individualized treatment effects (Chipman et al. (2010); Wager & Athey (2017); Athey & Imbens (2016); Lu et al. (2017); Alaa & van der Schaar (2017); Porter et al. (2011); Johansson et al. (2016); Alaa et al. (2017); Louizos et al. (2017); Shalit et al. (2017)).”
- The Blessings of Multiple Causes (2018, methodology): “Louizos et al. (2017) use variational autoencoders to infer unobserved confounders from proxy variables, Shah and Meinshausen (2018) develop projection-based techniques for high-dimensional covari-1576 ance estimation under latent confounding, Frot, Nandy, and Maathuis (2019) use linear factor…”