Prof. Dr. Tobias Friedrich


Four Papers accepted at IJCAI

We are proud to announce that four papers by our group members were accepted at the International Joint Conference on Artificial Intelligence (IJCAI) on 19-25 August in Macao, China. The acceptance rate of the conference was only about 15%. The following paragraphs describe the results of the four papers.

Real-world networks have many interesting properties. For example, the network might exhibit a dynamic nature, where changes are happening in its structure over time, and at the same time, those changes are not governed by a central authority, but by multiple people where each one of them follows their own interest. These dynamic changes can be modelled with temporal graphs and the selfish agents can be modelled with network creation games. In the paperTemporal Network Creation Games, the authors combine these two established concepts to create a model accurately captures the aforementioned properties of real-world networks. They analyze basic properties of the model such as hardness of computing the best agents strategy, and the quality of the constructed graphs.

The second paper Strategic Resource Selection with Homophilic Agents deals with the strategic selection of resources by selfish agents. Instead of the normal resource selection setting, the agents are heterogenous and their utility depends on the fraction of same-type agents using the resource. The authors show the existence and polynomial-time computability of equilibria and give bounds on the Price of Anarchy and Stability under some natural assumptions. This paper is based on the master thesis of Jonathan Gadea Harder who recently also joined our group as a PhD student.

In the modern world, information is becoming increasingly valuable and a big part of an employee's day in a company is spent exchanging information via meetings. Unfortunately, a lot of employees are reporting that they are becoming more and more unproductive due to spending time on meetings that are not really relevant to them. The goal of the paper Minimizing Reachability Times on Temporal Graphs via Shifting Labels is to examine how networks of information can be reconfigured such that the important information that needs to be disseminated into the network is spread as fast as possible with minimal number of changes to the schedule. The authors model such a network via temporal graphs where information between the nodes can only be exhanged in specific times steps. They show that such optimizations are generally computationally hard unless we study restricted classes of graphs or we bound some of the parameters of the network.

The last paper Continuous Schelling Games is based on the bachelor thesis of Jonas Schmidt and analyses population segregation caused by agents considering non-categorical attributes along a continuous scale, e.g. income, and extensively studies three natural models. The authors shed light on the existence and computation of equilibria, provide bounds on the Price of Anarchy and Stability, and insightful simulation results comparing the models.

  • Strategic Resource Select... - Download
    Gadea Harder, Jonathan; Krogmann, Simon; Lenzner, Pascal; Skopalik, Alexander Strategic Resource Selection with Homophilic AgentsInternational Joint Conference on Artificial Intelligence (IJCAI) 2023: 2701–2709
  • Minimizing Reachability T... - Download
    Deligkas, Argyrios; Eiben, Eduard; Skretas, George Minimizing Reachability Times on Temporal Graphs via Shifting LabelsInternational Joint Conference on Artificial Intelligence (IJCAI) 2023: 5333–5340
  • Temporal Network Creation... - Download
    Bilò, Davide; Cohen, Sarel; Friedrich, Tobias; Gawendowicz, Hans; Klodt, Nicolas; Lenzner, Pascal; Skretas, George Temporal Network Creation GamesInternational Joint Conference on Artificial Intelligence (IJCAI) 2023: 2511–2519
  • Schelling Games with Cont... - Download
    Bilò, Davide; Bilò, Vittorio; Döring, Michelle; Lenzner, Pascal; Molitor, Louise; Schmidt, Jonas Schelling Games with Continuous TypesInternational Joint Conference on Artificial Intelligence (IJCAI) 2023: 2520–2527