What about non-Gaussian?


Description
In many engineering problems, Gaussian noise is assumed. Sometimes for good reasons, sometimes for convenience. In information theory, another important property of Gaussian distributions, is that they maximize the entropy under a variance constraint. However, in many multi-user problems, this property is not sufficient to figure out the capacity achieving inputs. Indeed it could be helpful to use non-Gaussian input distributions. But non-Gaussian input distributions are hard to analyze over Gaussian noise channels. In [1], a technique using Hermite polynomials has been proposed for that purpose. The Goal of the project is to study this technique and its application to some problem.


Perequisit
Information theory, probability, analysis. The project can be approached from a more theoretic or applied persepctive.

[1] E. Abbe, L. Zheng, “Coding Along Hermite Polynomials for Gaussian Noise Channels”, ISIT 2009, Seoul. (Email if interested in the paper.)


Supervisor
Dr. Emmanuel Abbe, LCM, Office INR 130, tel 35657, emmanuel.abbe@epfl.ch


Professor
Prof. Bixio Rimoldi, LCM

back to projects menu

Last modified:: %2009/%09/%17 %15:%Sep