BACKSHIFT: Learning causal cyclic graphs from unknown shift interventions

By Dominik Rothenhäusler, Christina Heinze, Jan Peters, Nicolai Meinshausen. 2015

89 citations recorded by the source corpus.

We propose a simple method to learn linear causal cyclic models in the presence of latent variables. The method relies on equilibrium data of the model recorded under a specific kind of interventions ("shift interventions"). The location and strength of these interventions do not have to be known and can be estimated from the data. Our method, called BACKSHIFT, only uses second moments of the data and performs simple joint matrix diagonalization, applied to differences between covariance matrices. We give a sufficient and necessary condition for identifiability of the system, which is fulfilled almost surely under some quite general assumptions if and only if there are at least three distinct experimental settings, one of which can be pure observational data. We demonstrate the performance on some simulated data and applications in flow cytometry and financial time series.

Subfields: Artificial Intelligence, Molecular Biology, Signal Processing

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

In Molecular Biology

In Signal Processing

Citations in the corpus

Learning Neural Causal Models from Unknown Interventions (2019, background): “Rothenhäusler et al. (2015) have investigated the case of additive shift interventions, while Eaton & Murphy (2007a) have examined the situation where the targets of experimental interventions are imperfect or uncertain.”
Causal Structure Learning (2017, methodology): “All methods are called through the interface oﬀered by the CompareCausalNetworks package which depends on the packages backShift (Heinze-Deml, 2017), bnlearn (Scutari, 2010) and pcalg (Kalisch et al., 2012) for the code of the considered methods.”
D’ya Like DAGs? A Survey on Structure Learning and Causal Discovery (2021, background): “, CCD [196], backshift [204], CCI [225]), there are relatively few such continuous optimization approaches.”
Joint Causal Inference from Multiple Contexts (2016, methodology): “The latter approach is commonly used by constraint-based methods for causal discovery from multiple data sets (e.g., Tillman, 2009; Claassen and Heskes, 2010; Tillman and Spirtes, 2011; Hyttinen et al., 2014; Triantafillou and Tsamardinos, 2015; Rothenhäusler et al., 2015; Forré and Mooij, 2018).”
Ancestral Causal Inference (2016, unknown): “In contrast to likelihoodbased approaches like [21, 5, 15, 19], in our approach we do not need to model the interventions quantitatively.”
Permutation-Based Causal Structure Learning with Unknown Intervention Targets (2019, methodology): “A diﬀerent approach to this problem is given by the invariant causal inference framework (Mein-shausen et al., 2016; Rothenh¨ausler et al., 2015; Ghas-sami et al., 2017).”
Causal Discovery from Soft Interventions with Unknown Targets: Characterization and Learning (2020, background): “Another approach described in [23] learns cyclic causal graphs assuming linearity from unknown shift interventions, which is a speciﬁc type of soft interventions.”
Structural Agnostic Modeling: Adversarial Learning of Causal Graphs (2018, methodology): “…assumption and handle hidden confounders, e.g. by introducing statistical dependencies between the noise variables attached to different variables (Rothenhäusler et al., 2015), or creating shared noise variables (Janzing and Schölkopf, 2018) or proxies of confounders (Wang and Blei, 2021).”
SAM: Structural Agnostic Model, Causal Discovery and Penalized Adversarial Learning (2018, methodology): “Indeed, by introducing a noise co-variance matrix with learnable parameters (Rothenh¨ausler et al., 2015), one can estimate hidden confounding effects at a minimal computational cost.”