Applications of the Griffiths-Kelly-Sherman correlation inequality to LDPC codes

Speaker: Nicolas Macris


It will be explained how a correlation inequality of statistical mechanics can be applied to the theory of low density parity check codes. Two applications will be considered.

The first one yields a proof that the growth rate can be computed exactly by iterative methods at least on the interval where it is concave. The second concerns sharp bounds for the “generalised EXIT curve” in the case of communication through a noisy gaussian channel.