On concentration of measures for LDPC code ensembles

摘要

This work considers the concentration of measures for low-density parity-check (LDPC) code ensembles. The two results derived in this paper follow from Azuma''s inequality for Doob martingales with bounded differences. The first result is a tightened concentration inequality for the conditional entropy (originally derived by Méasson et al.), and the second result is a concentration inequality for the cardinality of the fundamental systems of cycles of a bipartite graph from the ensemble.