Invited Speakers
Invited Speakers
VU Amsterdam
Talk: Influence maximization in (social) networks
Abstract: Online social networks have become crucial communication channels. Millions of people participate in such networks including entities with commercial interests such as companies. The latter increasingly incorporate campaigns promoted via social networks into their marketing mix. Fundamental problems that arise in quantitative social network analysis in the context of (viral) marketing include the identification of influential network nodes that may trigger a large information propagation cascade and the identification of (homogeneous) communities. In this talk we focus on the former problem which is typically referred to as the influence maximization problem. We first provide an overview over considered problem variants and recently proposed exact algorithms for this problem. In the second part of the talk, we motivate several extensions that have been recently introduced. These include the (explicit) consideration of individuals who view content but do not forward it and alternative objective functions. We discuss a generic mixed-integer non-linear programming formulation that incorporates these aspects and an exact linearization based on generalized Benders decomposition. Finally, we report results from a computational study showing that the new problem variants can lead to more effective marketing campaigns and discuss topics for future research.
Fraunhofer ITWM, TU Kaiserslautern
Talk: Robust optimization: Concepts, results and applications in public transportation networks
Abstract: Most real-world problems contain parameters which are not known at the time a decision is to be made. Data may not be measurable in the precision needed or may depend on future developments. An optimal solution which does not take such an uncertainty into account often becomes bad or even infeasible for the scenario which is finally realized.
This is, e.g., true for optimal routes or for timetables in public transport which easily become infeasible if delays occur. In robust optimization one specifies the uncertainty as a scenario set. Classical robust optimization aims at finding a solution which is feasible for all scenarios. Such a solution comes with a high price: A robust route or a robust timetable would have much too long nominal traveling times. The concept is hence not suitable for many applications. There exist less conservative robustness concepts. In the first part of this talk, several such definitions will be shown. Two of them will be discussed in more detail: Light robustness and a scenario-based approach to recovery robustness.
The second part of the talk goes one step further: How to handle uncertain optimization problems in which more than one objective function is to be considered? This yields a robust multi-objective optimization problem, a class of problems only recently introduced.
Concepts on how to define robust Pareto solutions will be developed. Mathematical properties will be derived as well as approaches on how to compute robust efficient solutions.
All concepts will be illustrated at routing and planning problems in public transportation.
Syracuse University
Talk: The Current and Future Waves of Interdiction Models and Algorithms
Abstract: This presentation will consists of three parts. The first part will introduce concepts related to network interdiction and robust optimization over networks, with an eye toward describing the truly important work performed in this field in the 1960s and 1970s that could become increasingly important in future studies. The second part will discuss a selection of contemporary interdiction studies, especially those related to assumptions on information asymmetry, stochasticity, players’ risk preferences, and learning. The third part will cover a sampling of some fascinating new applications and problem variations that are currently being studied, with an eye toward forecasting what the next wave of research in this field might address.