Workshop Title: “Your model is wrong: robustness and misspecification in probabilistic modeling”

Location: Virtual, at NeurIPS 2021


Probabilistic modeling is a foundation of modern data analysis — due in part to the flexibility and interpretability of these methods — and has been applied to numerous application domains, such as the biological sciences, social and political sciences, engineering, and health care. However, any probabilistic model relies on assumptions that are necessarily a simplification of complex real-life processes; thus, any such model is inevitably misspecified in practice. In addition, as data set sizes grow and probabilistic models become more complex, applying a probabilistic modeling analysis often relies on algorithmic approximations, such as approximate Bayesian inference, numerical approximations, or data summarization methods. Thus in many cases, approximations used for efficient computation lead to fitting a misspecified model by design (e.g., variational inference). Importantly, in some cases, this misspecification leads to useful model inferences, but in others it may lead to misleading and potentially harmful inferences that may then be used for important downstream tasks for, e.g., making scientific inferences or policy decisions.

The workshop will bring together researchers focused on methods, applications, and theory to outline some of the core problems in specifying and applying probabilistic models in modern data contexts along with current state-of-the-art
solutions. Participants will leave the workshop with (i) exposure to recent advances in the field, (ii) an idea of the current major challenges in the field, and (iii) an introduction to methods meeting these challenges.

Confirmed Invited speakers: Chris Holmes (Oxford University), Jonathan Huggins (Boston University), Ilse Ipsen (North Carolina State University), Lester Mackey (Microsoft Research), Andres Masegosa (University of Almería), and Yixin Wang (University of California Berkeley)

Call for papers:

We invite submissions to the NeurIPS 2021 Workshop “Your model is wrong: robustness and misspecification in probabilistic modeling.” We welcome submissions that identify problems and potential solutions regarding misspecification, approximation, robustness, and reliability of probabilistic inference on topics including but not limited to:

– Nonparametric Bayesian models

– Probabilistic surrogate modeling

– Approximate Bayesian inference algorithms (e.g. MCMC and variational inference)

– Probabilistic numerical methods

– Data summarization for probabilistic models (e.g. sketching, subsampling, coresets)

– Missing data and data imputation

– Application areas highlight issues and harms arising from misspecification (e.g. health care, social sciences, economics, and scientific engineering applications)

Abstracts should be submitted via the CMT3 platform by 13 September 2021. Full submissions (no longer than 4 pages excluding references & acknowledgements) should be submitted via the CMT3 platform by 17 September 2021. Please see our website for formatting instructions and a link to the submission system. All submissions are non-archival.