Fermilab Computing Division

CS Document 6817-v1

Current Samplers and Hamiltonian Monte Carlo: Notes for Users of CosmoSIS

Document #:
CS-doc-6817-v1
Document type:
Technical Note
Submitted by:
Marc Paterno
Updated by:
Marc Paterno
Document Created:
26 Jul 2019, 13:42
Contents Revised:
26 Jul 2019, 13:42
Metadata Revised:
26 Jul 2019, 13:44
Viewable by:
  • Public document
Modifiable by:

Quick Links:
Latest Version

Abstract:
The purpose of this document is to explore the mathematical and statistical background behind sampling methods implemented in CosmoSIS. The document is broadly split up into four sections: Random-walk Metropolis Hastings, Hamiltonian Monte Carlo, Gradient-free Hamiltonian Monte Carlo and MultiNest. Random-walk Metropolis Hastings and MultiNest were already in use in CosmoSIS at the time when this document was written, while the purpose of this project was in part to determine the viability of Hamiltonian Monte Carlo as a sampling method in CosmoSIS. These notes are meant to be self-contained but are not comprehensive or entirely rigorous. Further reading on selected topics will be highlighted throughout the notes.
Files in Document:
Authors:
Notes and Changes:
The actual author is Scott Ziegler, who is not registered in a way that allows proper accreditation.
DocDB Home ]  [ Search ] [ Authors ] [ Events ] [ Topics ]

DocDB Version 8.8.9, contact Document Database Administrators