posterior-landscape: morphological feature finding in two-dimensional probability densities
In the authors' words
We present posterior-landscape, a Python framework to identify and characterize morphological features in two-dimensional probability densities. The framework combines concepts from topological data analysis, statistical density-ridge estimation, and scale-space image analysis. It extracts localized concentrations, valleys, ridges, and changes of slope; for Bayesian reconstructions, it propagates uncertainty in feature recovery, geometry, and correspondence across posteriors. The framework also connects two-dimensional structures to features in one-dimensional marginals, and quantifies their correlation with other inferred parameters. We validate the method on controlled densities, and demonstrate its use on gravitational-wave population reconstructions, where it provides a systematic way to identify and interpret population features and assess which ones carry information relevant for cosmology. The resulting catalogues also offer a common morphological representation for future comparisons between predictions and population reconstructions.
Appeared: Friday, October 30. arXiv. Preprint, not yet peer-reviewed.
Authors' comment: 15 pages (41 total), 8 figures (10 total). Code available at https://github.com/Mik3M4n/posterior-landscape