Reliable Communication of Highly Distributed Information

Abstract

Shannon’s theory of information [1] and subsequent generalizations to multiple users (for a survey see [2]) consider the situation of a small number of users each with an unlimited amount of information. The users communicate over a noisy channel with the goal of exchanging their information reliably. Here, we consider a complementary model. We assume a very large number of users, each with a small amount of information. We also assume that the communication takes place over a noisy channel but assume that the goal of the users is to compute a function reliably. This highly distributed information model is motivated by problems of decision making in a network. The users could be either a large number of processors, human beings, or simply the components of a logic circuit. In all cases, the noise is an inevitable physical limitation.

See contribution by Gallager in Chapter VI for more on this.