Constraint-based Causal Discovery: Conflict Resolution with Answer Set Programming

Recent approaches to causal discovery based on Boolean satisfiability solvers have opened new opportunities to consider search spaces for causal models with both feedback cycles and unmeasured confounders. However, the available methods have so far not been able to provide a principled account of how to handle conflicting constraints that arise from statistical variability. Here we present a new approach that preserves the versatility of Boolean constraint solving and attains a high accuracy despite the presence of statistical errors. We develop a new logical encoding of (in)dependence constraints that is both well suited for the domain and allows for faster solving. We represent this encoding in Answer Set Programming (ASP), and apply a state-of-the-art ASP solver for the optimization task. Based on different theoretical motivations, we explore a variety of methods to handle statistical errors. Our approach currently scales to cyclic latent variable models with up to seven observed variables and outperforms the available constraint-based methods in accuracy.

[1]  Frederick Eberhardt,et al.  Discovering Cyclic Causal Models with Latent Variables: A General SAT-Based Procedure , 2013, UAI.

[2]  Dimitris Margaritis,et al.  EFFICIENT MARKOV NETWORK DISCOVERY USING PARTICLE FILTERS , 2009, Comput. Intell..

[3]  Z. Jane Wang,et al.  Controlling the False Discovery Rate of the Association/Causality Structure Learned with the PC Algorithm , 2009, J. Mach. Learn. Res..

[4]  Jukka Corander,et al.  Learning Chordal Markov Networks by Constraint Satisfaction , 2013, NIPS.

[5]  Ilkka Niemelä,et al.  Logic programs with stable model semantics as a constraint programming paradigm , 1999, Annals of Mathematics and Artificial Intelligence.

[6]  Wei Luo,et al.  Mind change optimal learning of Bayes net structure from dependency and independency data , 2010, Inf. Comput..

[7]  Peter Spirtes,et al.  Learning equivalence classes of acyclic models with latent and selection variables from multiple datasets with overlapping variables , 2011, AISTATS.

[8]  Toby Walsh,et al.  Handbook of Satisfiability: Volume 185 Frontiers in Artificial Intelligence and Applications , 2009 .

[9]  Jiji Zhang,et al.  Adjacency-Faithfulness and Conservative Causal Inference , 2006, UAI.

[10]  Peter Bühlmann,et al.  Causal Inference Using Graphical Models with the R Package pcalg , 2012 .

[11]  J. Koster.  Marginalizing and conditioning in graphical models , 2002 .

[12]  Vincenzo Lagani,et al.  Towards Integrative Causal Analysis of Heterogeneous Data Sets and Studies , 2012, J. Mach. Learn. Res..

[13]  P. Spirtes,et al.  Causation, prediction, and search , 1993 .

[14]  Radford M. Neal.  On Deducing Conditional Independence from d-Separation in Causal Graphs with Feedback (Research Note) , 2000, J. Artif. Intell. Res..

[15]  Marius Thomas Lindauer,et al.  Potassco: The Potsdam Answer Set Solving Collection , 2011, AI Commun..

[16]  Tom Heskes,et al.  A Bayesian Approach to Constraint Based Causal Inference , 2012, UAI.

[17]  David Danks,et al.  Integrating Locally Learned Causal Structures with Overlapping Variables , 2008, NIPS.

[18]  David Maxwell Chickering,et al.  Optimal Structure Identification With Greedy Search , 2002, J. Mach. Learn. Res..

[19]  Peter Spirtes,et al.  Directed Cyclic Graphical Representations of Feedback Models , 1995, UAI.

[20]  Milan Studený.  Bayesian Networks from the Point of View of Chain Graphs , 1998, UAI.

[21]  Timo Soininen,et al.  Extending and implementing the stable model semantics , 2000, Artif. Intell..

[22]  J. Pearl.  Causality: Models, Reasoning and Inference , 2000 .

[23]  P. Spirtes,et al.  Ancestral graph Markov models , 2002 .

[24]  Rina Dechter,et al.  Identifying Independencies in Causal Graphs with Feedback , 1996, UAI.

[25]  Ioannis G. Tollis,et al.  Learning Causal Structure from Overlapping Variable Sets , 2010, AISTATS.

[26]  David Heckerman,et al.  Learning Gaussian Networks , 1994, UAI.

[27]  Ioannis Tsamardinos,et al.  Constraint-based causal discovery from multiple interventions over overlapping variable sets , 2014, J. Mach. Learn. Res..

[28]  Gregory F. Cooper,et al.  A Bayesian Method for the Induction of Probabilistic Networks from Data , 1992 .