Undecidability of approximating the capacity of time-invariant Markoff channel with feedback, and non-existence of linear finite-letter conditional mutual information character

04/16/2018
by   Mukul Agarwal, et al.
0

It is proved that approximating, within an additive constant, the capacity of time-invariant Markoff channels with feedback can be posed as a problem which is undecidacable. Then, a definition for the capacity region, which we call the linear finite-letter conditional mutual information characterization is provided, and it is proved that evaluating the feasibility of achievability of a certain rate for this characterization is a decidable problem if Schanuel's conjecture is true. Thus, assuming Schanuel's conjecture is true, there is no linear finite-letter mutual information characterization for the capacity of a time-invariant Markoff channel with feedback.

READ FULL TEXT

Please sign up or login with your details

Forgot password? Click here to reset
Success!
Error Icon An error occurred

Sign in with Google

×

Use your Google Account to sign in to DeepAI

×

Consider DeepAI Pro