Benefits of Rate-Sharing for Distributed Hypothesis Testing

TL;DR

This paper analyzes distributed binary hypothesis testing with a shared and individual communication links, demonstrating how rate-sharing can improve error exponents and revealing tradeoffs between decision centers.

Contribution

It characterizes the optimal exponents region under expected rate constraints and simplifies the analysis for the case with only a shared link.

Findings

Rate-sharing improves error exponents compared to maximum rate constraints.

The exponents region shows a tradeoff between the two decision centers.

Simplified expressions for the exponents region in the single shared link case.

Abstract

We study distributed binary hypothesis testing with a single sensor and two remote decision centers that are also equipped with local sensors. The communication between the sensor and the two decision centers takes place over three links: a shared link to both centers and an individual link to each of the two centers. All communication links are subject to expected rate constraints. This paper characterizes the optimal exponents region of the type-II error for given type-I error thresholds at the two decision centers and further simplifies the expressions in the special case of having only the single shared link. The exponents region illustrates a gain under expected rate constraints compared to equivalent maximum rate constraints. Moreover, it exhibits a tradeoff between the exponents achieved at the two centers.