Communication over Individual Channels -- a general framework

Abstract: We consider the problem of communicating over a channel for which no mathematical model is specified, and the achievable rates are determined as a function of the channel input and output sequences known a-posteriori, without assuming any a-priori relation between them. In a previous paper we have shown that the empirical mutual information between the input and output sequences is achievable without specifying the channel model, by using feedback and common randomness, and a similar result for real-valued input and output alphabets. In this paper, we present a unifying framework which includes the two previous results as particular cases. We characterize the region of rate functions which are achievable, and show that asymptotically the rate function is equivalent to a conditional distribution of the channel input given the output. We present a scheme that achieves these rates with asymptotically vanishing overheads.