Scaling possibilistic IMs to high-dimensional settings

Develop scalable computational and methodological strategies that enable possibilistic inferential models to operate effectively in high-dimensional parameter spaces while maintaining strong validity and achieving statistical efficiency.

Background

The paper’s validity results are finite-sample and dimension-agnostic, but high-dimensional applications pose computational and efficiency challenges.

The conclusion calls for advances that combine optimization and Monte Carlo techniques and incorporate regularization to address both statistical and computational scaling.