Planet-GLLiM: software for scalable Bayesian analysis of multidimensional data in astrophysics - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2022

Planet-GLLiM: software for scalable Bayesian analysis of multidimensional data in astrophysics

Abstract

Planet-GLLiM is a software to handle Bayesian inverse problems in the context of physical models inversion in planetary remote sensing. It is built in a computationally efficient way and can easily handle situations where the signals to be inverted present a moderately high number of dimensions and are in large number. The implemented model is based on a tractable inverse regression approach which has the advantage to produce full probability distributions as approximations of the target posterior distributions. In addition to provide confidence indices on the predictions, these distributions allow a better exploration of inverse problems when multiple equivalent solutions exist. These distributions can also be used for further refined predictions using importance sampling, while also providing a way to carry out uncertainty level estimation if necessary. The approach shows interesting capabilities both in terms of computational efficiency and multimodal inference. In this contribution, we propose to briefly describe the main model components and to show a demo of the software on some planetary material.
No file

Dates and versions

hal-03883020 , version 1 (02-12-2022)

Identifiers

  • HAL Id : hal-03883020 , version 1

Cite

Stan Borkowski, Sylvain Douté, Florence Forbes, Samuel Heidmann. Planet-GLLiM: software for scalable Bayesian analysis of multidimensional data in astrophysics. Journées de l'Action Spécifique Numérique Astrophysique (ASNUM 2022), Dec 2022, Lyon, France. ⟨hal-03883020⟩
59 View
0 Download

Share

Gmail Facebook X LinkedIn More