Abstract

Shannon’s channel capacity is a first-order asymptotic quantity equal to the supremum of rates at which a decaying error probability can be achieved as the block length increases. Second-order coding rates provide a more refined indication of the system performance by quantifying the convergence speed of the coding rate to capacity for a given target error probability. This talk will review well-known results on the second-order asymptotics for point-to-point communication, and then overview some recent developments in network information theory problems. Achievability results will be provided for the multiple-access channel (MAC) and Gel’fand-Pinsker channel, and conclusive second-order asymptotic results (i.e. matching achievability and converse) will be provided for two Gaussian settings: the MAC with degraded message sets, and dirty paper coding.