On small neighborhoods of the capacity-achieving input distributions, the decrease of the mutual information with the distance to the capacity-achieving input distributions is bounded below by a linear function of the square of the distance to the capacity-achieving input distributions for all channels with (possibly multiple) linear constraints and finite input sets using an identity due to Tops{\o}e and Pinsker's inequality. Counter examples demonstrating non-existence of such a quadratic bound are provided for the case of infinite many linear constraints and the case of infinite input sets. Using a Taylor series approximation, rather than Pinsker's inequality, the exact characterization of the slowest decrease of the mutual information with the distance to the capacity-achieving input distributions is determined on small neighborhoods of the capacity-achieving input distributions. Analogous results are established for classical-quantum channels whose output density operators are defined on a separable Hilbert spaces. Implications of these observations for the channel coding problem and applications of the proof technique to related problems are discussed.
翻译:暂无翻译