Typical Error Exponents: A Dual Domain Derivation

This paper shows that the probability that the error exponent of a given code randomly generated from a pairwise-independent ensemble is smaller than a lower bound on the typical random-coding exponent tends to zero as the codeword length tends to infinity. This lower bound is known to be tight for i.i.d. ensembles over the binary symmetric channel and for constant-composition codes over memoryless channels. Our results recover both as special cases and remain valid for arbitrary alphabets, arbitrary channels—for example finite-state channels with memory—, and arbitrary pairwise-independent ensembles. We specialize our results to the i.i.d., constant-composition and cost-constrained ensembles over discrete memoryless channels and to ensembles over finite-state channels.