Title: When is structured coding useful in goal-oriented communication? A case study of distributed function computation

Time: 10:30-11:30 CET, Fri Mar 06, 2026

Location: Lehrstuhl für Nachrichtentechnik, TU Dortmund, Germany

Local Host: Onur Günlü, onur.guenlue@tu-dortmund.de

Abstract: The goal oriented communication seeks to improve communication efficiency only by send

relevant information that is needed to accomplish a given task at the destination. In this talk,

we review the goal oriented communication from a perspective of distributed function computation

problem. Particular focus is on whether the achievable rate region can be improved by the structured coding or not. If the destination can recover the entire data, then a function value can be computed at the destination. Thus, the optimal rate region for distributed compression, the Slepian-Wolf region, is a trivial achievability bound for the distributed function computation. In 1979, Korner and Marton showed that, for the module-sum problem, a broader rate region can be achieved by using a structured coding. In 1987, Han and Kobayashi derived the necessary and sufficient condition on functions such that the Slepian-Wolf region cannot be improved for any sources having positive support. For more than two encoders, Han and Kobayashi derived sufficient conditions and necessary conditions such that the Slepian-Wolf region cannot be improved. An improved sufficient condition was further derived by Watanabe. However, the classification problem for more than two encoders is unsolved. 

Since the classification of functions is solved for two encoders case, our next question is to find out 

the condition on sources such that the Slepian-Wolf region can be improved for a given function, such as the modulo-sum. A sufficient condition on optimality of Slepian-Wolf coding for the modulo-sum is known. By a hybrid use of random coding and structured coding, Ahlswede and Han derived an achievable rate region that improves upon the Slepian-Wolf region and the Korner-Marton region. However, the sum rate may not be improved by the Ahlwede-Han region. More recently, it was numerically demonstrated that the multi-letter Ahlswede-Han region strictly improves upon the Slepian-Wolf region if the aforementioned sufficient condition is violated. 

In this talk, we review these developments of the distributed function computation problem and provide some future perspective.

About the Speaker: https://sites.google.com/site/shunwatanabeshomepage/shun-watanabes-homepage?authuser=0

Shun Watanebe of the Tokyo University of Agriculture and Technology, Japan has visited the Lehrstuhl für Nachrichtentechnik at TU Dortmund University in March 2026.