Sculpting priors

TL;DR

This paper explores the design of Bayesian priors tailored for detecting targets with known spectral signatures in cluttered backgrounds, aiming to outperform traditional non-Bayesian methods like the GLRT.

Contribution

It introduces a novel approach to sculpting Bayesian priors that consistently surpass the performance of the standard GLRT in target detection tasks.

Findings

Sculpted priors improve detection accuracy in cluttered environments.

The method outperforms traditional GLRT in various scenarios.

Bayesian priors can be optimized for specific detection tasks.

Abstract

Bayesian priors are investigated for detecting targets of known spectral signature (but unknown strength) in cluttered backgrounds. A specific problem is the construction (or ``sculpting'') of a Bayesian prior that uniformly outperforms its non-Bayesian counterpart, the nominally sub-optimal but widely used Generalized Likelihood Ratio Test (GLRT).