Partition function approach to non-Gaussian likelihoods: macrocanonical partitions and replicating Markov-chains

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Herzog, Maximilian Philipp, von Campe, Heinrich, Kuntz, Rebecca Maria, Röver, Lennart, Schäfer, Björn Malte
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912094457167872
author Herzog, Maximilian Philipp
von Campe, Heinrich
Kuntz, Rebecca Maria
Röver, Lennart
Schäfer, Björn Malte
author_facet Herzog, Maximilian Philipp
von Campe, Heinrich
Kuntz, Rebecca Maria
Röver, Lennart
Schäfer, Björn Malte
contents Monte-Carlo techniques are standard numerical tools for exploring non-Gaussian and multivariate likelihoods. Many variants of the original Metropolis-Hastings algorithm have been proposed to increase the sampling efficiency. Motivated by Ensemble Monte Carlo we allow the number of Markov chains to vary by exchanging particles with a reservoir, controlled by a parameter analogous to a chemical potential $μ$, which effectively establishes a random process that samples microstates from a macrocanonical instead of a canonical ensemble. In this paper, we develop the theory of macrocanonical sampling for statistical inference on the basis of Bayesian macrocanonical partition functions, thereby bringing to light the relations between information-theoretical quantities and thermodynamic properties. Furthermore, we propose an algorithm for macrocanonical sampling, $\texttt{Avalanche Sampling}$, and apply it to various toy problems as well as the likelihood on the cosmological parameters $Ω_m$ and $w$ on the basis of data from the supernova distance redshift relation.
format Preprint
id arxiv_https___arxiv_org_abs_2311_16218
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Partition function approach to non-Gaussian likelihoods: macrocanonical partitions and replicating Markov-chains
Herzog, Maximilian Philipp
von Campe, Heinrich
Kuntz, Rebecca Maria
Röver, Lennart
Schäfer, Björn Malte
Cosmology and Nongalactic Astrophysics
Instrumentation and Methods for Astrophysics
Data Analysis, Statistics and Probability
Monte-Carlo techniques are standard numerical tools for exploring non-Gaussian and multivariate likelihoods. Many variants of the original Metropolis-Hastings algorithm have been proposed to increase the sampling efficiency. Motivated by Ensemble Monte Carlo we allow the number of Markov chains to vary by exchanging particles with a reservoir, controlled by a parameter analogous to a chemical potential $μ$, which effectively establishes a random process that samples microstates from a macrocanonical instead of a canonical ensemble. In this paper, we develop the theory of macrocanonical sampling for statistical inference on the basis of Bayesian macrocanonical partition functions, thereby bringing to light the relations between information-theoretical quantities and thermodynamic properties. Furthermore, we propose an algorithm for macrocanonical sampling, $\texttt{Avalanche Sampling}$, and apply it to various toy problems as well as the likelihood on the cosmological parameters $Ω_m$ and $w$ on the basis of data from the supernova distance redshift relation.
title Partition function approach to non-Gaussian likelihoods: macrocanonical partitions and replicating Markov-chains
topic Cosmology and Nongalactic Astrophysics
Instrumentation and Methods for Astrophysics
Data Analysis, Statistics and Probability
url https://arxiv.org/abs/2311.16218