Half-time seminar: Algorithms and benchmarking in uncertain networks

Abstract:

In uncertain networks, the likelihood of existence of edges is represented by probabilities. Uncertainty is an inherent property when modelling a system as a network due to the randomness of the system, inaccuracy of measurements, or their interpretation. Creating measures and algorithms on uncertain networks presents unique challenges compared to deterministic networks, because probabilistic edges require accounting for many possible network configurations.

This seminar is about clustering on uncertain networks, and we focus on two parts. The first part introduces efficient computation of expected modularity. Modularity is one of the most widely used measures for evaluating the quality of clustering in networks. In the second part, we introduce our benchmarking to study the impact of topological structure, probability distribution, and the interpretation of probabilities as weights on clustering algorithms on uncertain networks.