Speaker
Babak Mamandipoor
Title
Capacity-Achieving Distributions of Gaussian Multiple Access Channel With Peak Constraints
Abstract
Characterizing the probability distribution function for the input of a communication channel that achieves the maximum possible data rate, is one of the most fundamental problems in the field of information theory. In his groundbreaking paper, Shannon showed that the capacity of a point-to-point additive white Gaussian noise channel under an average power constraint at the input, is achieved by Gaussian distribution. Although imposing a limitation on the peak of the channel input is also very important in modeling the communication system more accurately, it has gained much less attention in the past few decades. A rather unexpected result of Smith indicated that the capacity achieving distribution for an AWGN channel under peak constraint at the input is unique and discrete, possessing a finite number of mass points.
In this thesis, we study multiple access channel under peak constraints at the inputs of the channel. By extending Smith's argument to out multi-terminal problem we show that any point on the boundary of the capacity region of the channel is only achieved by discrete distributions with a finite number of mass points. Although we do not claim uniqueness of the capacity-achieving distributions, however, we show that only discrete distributions with a finite number of mass points can achieve points on the boundary of the capacity region.
First we deal with the problem of maximizing the sum-rate of a two user Gaussian MAC with peak constraints. It is shown that generating the code-books of both users according to discrete distributions with a finite number of mass points achieves the largest sum-rate in the network. After that we generalize our proof to maximize the weighted sum-rate of the channel and show that the same properties hold for the optimum input distributions. This completes the proof that the capacity region of a two-user Gaussian MAC is achieved by discrete input distributions with a finite number of mass points.
Supervisor
Amir Khandani