Ulrik Brandes

Last Name

Brandes

First Name

Ulrik

ORCID

0000-0002-1520-0430

Organisational unit

09610 - Brandes, Ulrik / Brandes, Ulrik

Show all metadata

Search Results

Publications 1 - 10 of 75

Quality Metrics for Information Visualization
Behrisch, Michael; Blumenschein, Michael; Kim, Nam Wook; et al. (2018)
Computer Graphics Forum
Graphframex: Towards systematic evaluation of explainability methods for graph neural networks
Amara, Kenza; Ying, Rex; Zhang, Zitao; et al. (2022)
Proceedings of Machine Learning Research ~ Proceedings of the First Learning on Graphs Conference
As one of the most popular machine learning models today, graph neural networks (GNNs) have attracted intense interest recently, and so does their explainability. Users are increasingly interested in a better understanding of GNN models and their outcomes. Unfortunately, today’s evaluation frameworks for GNN explainability often rely on few inadequate synthetic datasets, leading to conclusions of limited scope due to a lack of complexity in the problem instances. As GNN models are deployed to more mission-critical applications, we are in dire need for a common evaluation protocol of explainability methods of GNNs. In this paper, we propose, to our best knowledge, the first systematic evaluation framework for GNN explainability, considering explainability on three different “user needs”. We propose a unique metric that combines the fidelity measures and classifies explanations based on their quality of being sufficient or necessary. We scope ourselves to node classification tasks and compare the most representative techniques in the field of input-level explainability for GNNs. For the inadequate but widely used synthetic benchmarks, surprisingly shallow techniques such as personalized PageRank have the best performance for a minimum computation time. But when the graph structure is more complex and nodes have meaningful features, gradientbased methods are the best according to our evaluation criteria. However, none dominates the others on all evaluation dimensions and there is always a trade-off. We further apply our evaluation protocol in a case study for frauds explanation on eBay transaction graphs to reflect the production environment.
Efficient Generation of Large Random Networks
Batagelj, Vladimir; Brandes, Ulrik (2005)
Physical Review E
Plainware and Polychrome: Quantifying Perceptual Differences in Ceramic Classification Between Diverse Groups to Further a Strong Objectivity
Borck, Lewis; Athenstädt, Jan C.; Cheromiah, Lee Ann; et al. (2020)
Journal of Computer Applications in Archaeology
A common problem when classifying archaeological objects is a potential cultural bias of the person deciding on the classification system. These are existing concerns within archaeology and anthropology and have previously been discussed as an emic/etic divide, “folk” classifications, or objective versus subjective approaches. But who gets to decide what is objective is often a subjective endeavour. To examine if and how cultural perceptions bias classification systems, we use methods from the field of cultural domain analysis to quantify differences in perception of ceramic sherds between different groups of people, specifically archaeologists and Indigenous and non-Indigenous potters. For this study, we asked participants to arrange a set of 30 archaeological sherds on a canvas, then interviewed them following each sorting exercise. A geosocial analysis of the arrangements in this pilot study suggests that there are substantial differences in the criteria by which the sherds are sorted between the groups. In particular, the arrangements by the Indigenous potters showed a greater diversity in the selection of underlying attributes. Understanding our different perceptions towards the material we use to construct history is the first step towards approaching a strong objectivity and thus a less fraught and more culturally inclusive discipline.
Centrality in directed networks
Marmulla, Gordana; Brandes, Ulrik (2026)
Social Networks
The identification of important nodes in a network is a pervasive task in a variety of disciplines from sociology and bibliometry to geography and chemistry, and an ever growing number of centrality indices is proposed for this purpose. While such indices are often ad-hoc, preservation of the vicinal preorder has been identified as the core axiom shared by centrality rankings on undirected graphs. We extend this idea to directed graphs by defining vertex preorders based on directed neighborhood-inclusion criteria. While, for the undirected case, the vicinal preorder is total on threshold graphs and preserves all standard centrality indices, we show that our generalized preorders are total on certain subclasses of threshold digraphs. We thus provide a consistent formalization of the hitherto rather conceptual notions of radial, medial, and hierarchical centralities. Through the criteria different notions of centrality are distinguishable, as we exemplify with selected standard centrality indices.
Experiments on Graph Clustering Algorithms
Brandes, Ulrik; Gaertler, Marco; Wagner, Dorothea (2003)
Lecture Notes in Computer Science ~ Algorithms - ESA 2003
A promising approach to graph clustering is based on the intuitive notion of intra-cluster density vs. inter-cluster sparsity. While both formalizations and algorithms focusing on particular aspects of this rather vague concept have been proposed no conclusive argument on their appropriateness has been given. As a first step towards understanding the consequences of particular conceptions, we conducted an experimental evaluation of graph clustering approaches. By combining proven techniques from graph partitioning and geometric clustering, we also introduce a new approach that compares favorably.
Analysis and Visualization of Social Networks
Brandes, Ulrik; Wagner, Dorothea (2004)
Graph Drawing Software
Robustness of preferential-attachment graphs
Hasheminezhad, Rouzbeh; Brandes, Ulrik (2023)
Applied Network Science
The widely used characterization of scale-free networks as "robust-yet-fragile" originates primarily from experiments on instances generated by preferential attachment. According to this characterization, scale-free networks are more robust against random failures but more fragile against targeted attacks when compared to random networks of the same size. Here, we consider a more appropriate baseline by requiring that the random networks match not only the size but also the inherent minimum degree of preferential-attachment networks they are compared with. Under this more equitable condition, we can (1) prove that random networks are almost surely robust against any vertex removal strategy and (2) show through extensive experiments that scale-free networks generated by preferential attachment are not particularly robust against random failures. Finally, we (3) add experiments demonstrating that preferentially attaching to well-connected vertices does not enhance robustness at all.
Editors' Note
Wasserman, Stanley; Brandes, Ulrik (2022)
Network Science
We welcome our new editors and provide background on an unusual duo of articles in this issue.
Central Positions in Social Networks
Brandes, Ulrik (2020)
Lecture Notes in Computer Science ~ Computer Science - Theory and Applications
This contribution is an overview of our recent work on the concept of centrality in networks. Instead of proposing new centrality indices, providing faster algorithms, or presenting new rules for when an index can be classified as a centrality, this research shifts the focus to the more elementary question whether a node is in a more central position than another. Viewing networks as data on overlapping dyads, and defining the position of a node as the whole of its relationships to the rest of the network, we obtain a very general procedure for doing centrality analysis; not only on social networks but networks from all kinds of domains. Our framework further suggests a variety of computational challenges. © Springer Nature Switzerland AG 2020.

Publications 1 - 10 of 75

Ulrik Brandes

Last Name

First Name

ORCID

Organisational unit

Filters

Search Results