Abstract
The concept of reducing the required transmission rate for a given system through prediction, interpolation, or other such techniques loosely labeled as "data compression" is now well known. The problems in analyzing such systems by theoretical means are formidable in even the simplest situations due to the inherent nonlinear nature of the operations performed. This paper discusses these difficulties, presents some approximate and exact solutions, and suggests areas where further work is needed.