Feature Extraction for Universal Hypothesis Testing via Rank-constrained Optimization

TL;DR

This paper introduces a novel feature extraction framework for universal hypothesis testing, utilizing rank-constrained optimization to improve detection performance in large observation spaces with poorly modeled alternatives.

Contribution

It proposes a new rank-constrained optimization approach for feature extraction, supported by theoretical bounds and a gradient-based algorithm with convergence guarantees.

Findings

Bounds on the number of distinguishable distributions in an exponential family

Development of a gradient-based algorithm for the optimization problem

Proof of local convergence of the proposed algorithm

Abstract

This paper concerns the construction of tests for universal hypothesis testing problems, in which the alternate hypothesis is poorly modeled and the observation space is large. The mismatched universal test is a feature-based technique for this purpose. In prior work it is shown that its finite-observation performance can be much better than the (optimal) Hoeffding test, and good performance depends crucially on the choice of features. The contributions of this paper include: 1) We obtain bounds on the number of \epsilon distinguishable distributions in an exponential family. 2) This motivates a new framework for feature extraction, cast as a rank-constrained optimization problem. 3) We obtain a gradient-based algorithm to solve the rank-constrained optimization problem and prove its local convergence.