For any channel P Y|X the strong data processing constant is defined as the smallest number η KL\in [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).
Bibliographical notePublisher Copyright:
© 1963-2012 IEEE.
- Contraction coefficient
- Strong data processing inequality (SDPI)
- contraction coefficient