The Non-Regular CEO Problem

TL;DR

This paper analyzes the asymptotic behavior of the CEO problem for non-regular source distributions, revealing a $1/R^2$ convergence rate of the mean squared error distortion as the number of agents and total rate grow large.

Contribution

It introduces a new asymptotic regime for the non-regular CEO problem, bridging the gap between discrete and Gaussian cases, with a layered coding scheme and a novel converse bound.

Findings

Distortion converges at a rate of 1/R^2 for large number of agents and rates.

Achievability demonstrated via layered scalar quantization and distributed coding.

Converse established using Bayesian Chazan-Zakai-Ziv bound.

Abstract

We consider the CEO problem for non-regular source distributions (such as uniform or truncated Gaussian). A group of agents observe independently corrupted versions of data and transmit coded versions over rate-limited links to a CEO. The CEO then estimates the underlying data based on the received coded observations. Agents are not allowed to convene before transmitting their observations. This formulation is motivated by the practical problem of a firm's CEO estimating (non-regular) beliefs about a sequence of events, before acting on them. Agents' observations are modeled as jointly distributed with the underlying data through a given conditional probability density function. We study the asymptotic behavior of the minimum achievable mean squared error distortion at the CEO in the limit when the number of agents $L$ and the sum rate $R$ tend to infinity. We establish a $1/R^2$ convergence of the distortion, an intermediate regime of performance between the exponential behavior in discrete CEO problems [Berger, Zhang, and Viswanathan (1996)], and the $1/R$ behavior in Gaussian CEO problems [Viswanathan and Berger (1997)]. Achievability is proved by a layered architecture with scalar quantization, distributed entropy coding, and midrange estimation. The converse is proved using the Bayesian Chazan-Zakai-Ziv bound.