Core Concepts
This work analyzes the minimum achievable average Age of Incorrect Information (AoII) in the non-asymptotic regime, where the impact of feedback time instances for variable-length stop-feedback (VLSF) codes is investigated.
Abstract
The key highlights and insights of the content are:
The authors study the average Age of Incorrect Information (AoII) in the context of remote monitoring of a symmetric Markov source using variable-length stop-feedback (VLSF) coding.
They consider sources with small cardinality, where feedback is non-instantaneous, as the transmitted information and feedback may have comparable lengths.
The authors leverage recent results on the non-asymptotic achievable channel coding rate to derive optimal feedback sequences, i.e., the times of feedback transmissions, in terms of either AoII or delay.
The results showcase the impact of the feedback sequence and SNR on the AoII, revealing that a lower average delay does not necessarily correspond to a lower average AoII.
The authors formulate the optimization problem as a Markov decision process (MDP) and develop MDPs for both AoII-optimal and delay-optimal feedback sequences of VLSF codes. They also compute the delay-minimal periodic feedback sequences as a baseline reference.
Numerical results illustrate that delay-optimality does not necessarily imply AoII-optimality, and the structure of the feedback sequence plays a significant role, with periodic feedback sequences performing consistently close to the AoII-optimal.