On Cognitive Interference Networks

TL;DR

This paper analyzes the asymptotic sum-rate capacity of multi-user interference networks with side-information, providing bounds and characterizations of when full or near-full capacity pre-log can be achieved.

Contribution

It offers a complete characterization of the pre-log behavior in interference networks with message side-information, including bounds and conditions for full pre-log.

Findings

Complete characterization of networks with full pre-log

Bounds on the pre-log based on side-information

Conditions for achieving near-full pre-log

Abstract

We study the high-power asymptotic behavior of the sum-rate capacity of multi-user interference networks with an equal number of transmitters and receivers. We assume that each transmitter is cognizant of the message it wishes to convey to its corresponding receiver and also of the messages that a subset of the other transmitters wish to send. The receivers are assumed not to be able to cooperate in any way so that they must base their decision on the signal they receive only. We focus on the network's pre-log, which is defined as the limiting ratio of the sum-rate capacity to half the logarithm of the transmitted power. We present both upper and lower bounds on the network's pre-log. The lower bounds are based on a linear partial-cancellation scheme which entails linearly transforming Gaussian codebooks so as to eliminate the interference in a subset of the receivers. Inter alias, the bounds give a complete characterization of the networks and side-information settings that result in a full pre-log, i.e., in a pre-log that is equal to the number of transmitters (and receivers) as well as a complete characterization of networks whose pre-log is equal to the full pre-log minus one. They also fully characterize networks where the full pre-log can only be achieved if each transmitter knows the messages of all users, i.e., when the side-information is "full".