Reent Schlegel defended his PhD thesis
Reent Schlegel defended his PhD thesis

Reent Schlegel defended his PhD thesis


On Friday 21 April 2023 at 14:00, Reent Schlegel defended his PhD thesis "Coding for Privacy in Distributed Computing". The defence took place in the Høyteknologisenteret, Lille auditorium at UiB.

Main research findings

In a distributed computer network, multiple entities collaborate to solve a problem. In this way, we can achieve more than the sum of the parts: cooperation allows the problem to be solved more efficiently, and at the same time it becomes possible to solve problems that each individual unit cannot solve on its own. On the other hand, devices that take a very long time to complete their task can significantly increase the total computation time. This so-called straggler effect can occur as a result of random events such as memory access and tasks running in the background on the various devices. The straggler problem usually blocks the entire computation since all devices must wait for the slowest devices to finish. Furthermore, the sharing of data and sub-calculations between the various units can put a significant strain on the communication network. Especially in a wireless network where the devices have to share a single communication channel, for example in calculations along the edge of a network (so-called edge calculations) and in federated learning, communication often becomes the bottleneck. Last but not least, sharing data with untrusted devices raises privacy concerns. Someone who wants to use a distributed computer network may be skeptical about sharing personal data with other entities without adequately protecting sensitive information. This thesis studies how ideas from code theory can mitigate the straggler problem, increase the efficiency of communication and guarantee data protection in distributed computing. In particular, a method from the thesis achieves an 8% speed improvement for edge calculations compared to a recently proposed method. At the same time, the method safeguards data protection, while the method that the thesis compares with does not. Another method from the thesis is up to 18 times faster for some use cases for federated learning compared to previous methods in the literature.

Adjudication committee:

  • Prof. Dr. Giuseppe Caire, Technical University Berlin
  • Assistant Professor Dr. Basak Guler , University of California, Riverside
  • Associate Professor Dr. Nikolay Kaleyski
  • Associate Professor Michal Walicki


  • Dr. Eirik Rosnes, Simula UiB
  • Prof. Alexandre Graell in Amat, Chalmers University

Read moreat the UiB new doctorates page.