DiBS: Differentiable Bayesian Structure Learning

Lars Lorch, Jonas Rothfuss, Bernhard Schölkopf, Andreas Krause

Keywords: causality, structure learning, Bayesian networks, Bayesian inference

TL;DR: A fully differentiable method for joint Bayesian inference of graphs and parameters of general Bayesian networks

Abstract: Bayesian structure learning allows inferring Bayesian network structure from data while reasoning about the epistemic uncertainty---a key element towards enabling active causal discovery and designing interventions in real world systems. In this work, we propose a general, fully differentiable framework for Bayesian structure learning (DiBS) that operates in the continuous space of a latent probabilistic graph representation. Contrary to existing work, DiBS is agnostic to the form of the local conditional distributions and allows for joint posterior inference of both the graph structure and the conditional distribution parameters. This makes our formulation directly applicable to posterior inference of nonstandard Bayesian network models, e.g., with nonlinear dependencies encoded by neural networks. Using DiBS, we devise an efficient, general purpose variational inference method for approximating distributions over structural models. In evaluations on simulated and real-world data, our method significantly outperforms related approaches to joint posterior inference.

Code Of Conduct: I certify that all co-authors of this work have read and commit to adhering to the NeurIPS Statement on Ethics, Fairness, Inclusivity, and Code of Conduct.

Code: https://github.com/larslorch/dibs

Community Implementations: [ 3 code implementations](https://www.catalyzex.com/paper/arxiv:2105.11839/code)