Strong data processing constant is achieved by binary inputs

2020/09/16 by Ordentlich, Or, Polyanskiy, Yury
#FOS: Computer and information sciences #Information Theory (cs.IT)

paper · doi:10.48550/arxiv.2010.01987

Abstract

For any channel P_Y|X the strong data processing constant is defined as the smallest number η_KL∈[0,1] such that I(U;Y)≤ η_KL I(U;X) holds for any Markov chain U-X-Y. It is shown that the value of η_KL is given by that of the best binary-input subchannel of P_Y|X. The same result holds for any f-divergence, verifying a conjecture of Cohen, Kemperman and Zbaganu (1998).

Related

• Strong data-processing inequalities for channels and Bayesian networks
• Strong Data Processing Inequalities for Input Constrained Additive Noise Channels
• Tight Exponential Strong Converses for Lossy Source Coding with Side-Information and Distributed Function Computation
• Strong Converses using Typical Changes of Measures and Asymptotic Markov Chains
• Strong Data Processing Inequalities and their Applications to Reliable Computation