Density estimation on small datasets

Abstract

How might a smooth probability distribution be estimated, with accurately quantified uncertainty, from a limited amount of sampled data? Here we describe a field-theoretic approach that addresses this problem remarkably well in one dimension, providing an exact nonparametric Bayesian posterior without relying on tunable parameters or large-data approximations. Strong non-Gaussian constraints, which require a non-perturbative treatment, are found to play a major role in reducing distribution uncertainty. A software implementation of this method is provided.

arXiv commentIncludes main text (5 pages, 3 figures) and Supplemental Information (10 pages, 4 figures). Same as version 3 but with Feynman diagrams properly rendered

References in corpus (6)

• Observation of a new boson at a mass of 125 GeV with the CMS experiment at the LHC
• Information field theory for cosmological perturbation reconstruction and non-linear signal analysis
• Field Theories for Learning Probability Distributions
• Occam factors and model-independent Bayesian learning of continuous distributions
• Rapid and deterministic estimation of probability densities using scale-free field theories
• Unification of field theory and maximum entropy methods for learning probability densities

Cited by in corpus (5)

• Physical limit to concentration sensing in a changing environment
• A Data-Driven Technique Using Millisecond Transients to Measure the Milky Way Halo
• Causal, Bayesian, & Non-parametric Modeling of the SARS-CoV-2 Viral Load Distribution vs. Patient's Age
• Entropy Estimation via Uniformization
• Search for Diffuse Galactic Neutrinos with the Full ANTARES Telescope Dataset