Katharina A. ZWEIG
Universitat Rovira i Virgili, Spain
Finding the most versatile nodes in highly multidimensional data
The determination of the most central agents in complex networks is important
because they are responsible for a faster propagation of information, epidemics,
failures and congestion, among others. A challenging problem is to identify them in
networked systems characterized by different types of interactions, forming
interconnected multilayer networks. Here we describe a mathematical framework
that allows us to calculate centrality in such networks and rank nodes accordingly,
finding the ones that play the most central roles in the cohesion of the whole structure, bridging together different types of relations. These nodes are the most
versatile in the multilayer network. We investigate empirical interconnected
multilayer networks and show that the approaches based on aggregating—or
neglecting—the multilayer structure lead to a wrong identification of the most
versatile nodes, overestimating the importance of more marginal agents and
demonstrating the power of versatility in predicting their role in propagation
processes with applications in social networks, banking networks, etc.
NOKIA Bell Labs, UK
Good City Life
The corporate smart-city rhetoric is about efficiency, predictability, and security.“You’ll get to work on time; no queue when you go shopping, and you are safebecause of CCTV cameras around you”. Well, all these things make a cityacceptable, but they don’t make a city great. We are launching goodcitylife.org - aglobal group of like-minded people who are passionate about building technologieswhose focus is not necessarily to create a smart city but to give a good life to citydwellers. The future of the city is, first and foremost, about people, and thosepeople are increasingly networked. We will see how a creative use of networkgenerated data can tackle hitherto unanswered research questions. Can we rethinkexisting mapping tools? Is it possible to capture smellscapes of entire cities andcelebrate good odors? And soundscapes?
Indiana University, USA
He received his PhD in Theoretical Physics in 2000 at theDepartment of Physics of the University of Bielefeld,Germany, working on lattice gauge theories, percolationand phenomenology of heavy-ion collisions. He switchedto complexity science in 2004, and from 2005 till 2007 hehas been postdoctoral researcher at the School ofInformatics and Computing of Indiana University, workingin the group of Alessandro Vespignani. From 2007 till2011 he has been at ISI Foundation in Turin, Italy, first as research scientist then as ascientific leader. In 2011 he became Associate Professor in Complex Systems at theSchool of Science of Aalto University, Finland. He is currently full professor in theSchool of Informatics and Computing at Indiana University.
Community structure in complex networks
Complex systems typically display a modular structure, as modules are easier toassemble than the individual units of the system, and more resilient to failures. Inthe network representation of complex systems, modules, or communities, appearas subgraphs whose nodes have an appreciably larger probability to get connectedto each other than to other nodes of the network. In this talk I will address threefundamental questions: How is community structure generated? How to detect it?How to test the performance of community detection algorithms? I will show thatcommunities emerge naturally in growing network models favoring triadic closure, amechanism necessary to implement for the generation of large classes of systems,like e.g. social networks. I will discuss the limits of the most popular class ofclustering algorithms, those based on the optimization of a global quality function,like modularity maximization. Testing algorithms is probably the single mostimportant issue of network community detection, as it implicitly involves theconcept of community, which is still controversial. I will discuss the importance ofusing realistic benchmark graphs with built-in community structure, as well as therole of metadata.
Purdue University, USA
Jennifer Neville is the Miller Family Chair AssociateProfessor of Computer Science and Statistics at PurdueUniversity. She received her PhD from the University ofMassachusetts Amherst in 2006. She is currently anelected member of the AAAI Executive Council and shewas recently PC chair of the 9th ACM InternationalConference on Web Search and Data. In 2012, she wasawarded an NSF Career Award, in 2008 she was chosenby IEEE as one of "AI's 10 to watch", and in 2007 was selected as a member of theDARPA Computer Science Study Group. Her work, which includes more than 100publications with over 5000 citations, focuses on developing data mining andmachine learning techniques for complex relational and network domains, includingsocial, information, and physical networks.
The impact of network structure on relational machine learning
Network science focuses on analyzing network structure in order to understand keyrelational patterns in complex systems. In contrast, relational machine learningtypically conditions on the observed relations in a network, using them as a form ofinductive bias to constrain the space of dependencies considered by the models.While recent interest in these two fields has produced a large body of research onmodels of both network structure and relational data, there has been less attentionon the intersection of the two fields--specifically considering the impact of networkstructure on relational learning methods. Since many relational domains comprise asingle, large, partially-labeled network, many of the conventional assumptions inrelational learning are no longer valid and the network structure creates uniquestatistical challenges for learning and inference algorithms. This talk will outlinesome of the algorithmic and statistical challenges that arise due to partiallyobserved, large-scale networks, and describe methods for semi-supervised learning,latent-variable modeling, and sampling to address the challenges.
ETH Zurich, Switzerland
Spreading influence in social networks: From link-centric to nodecentric models
Epidemic spreading on complex networks is well studied because nodes follow arather simple dynamic. Thus, the focus is mostly on how the network topologyimpacts the spreading process. However, modeling the spread of, e.g., emotions inonline social networks requires us to have more refined models of the nodedynamics, to calculate cascades of spreading influence. We capture the nodedynamics by means of a data-driven modeling approach that allows us to test, and to calibrate, assumptions about the user behavior. In my talk, I present differentexamples of how to complement the topological perspective by a node-centricperspective that considers costs and benefits, emotional responses or informationprocessing of users.
Katharina A. ZWEIG
TU Kaiserslautern, Germany
Network analysis literacy: a socioinformatic approach
Why are there so many centrality indices? This is the question that puzzled mewhen I started into network analysis in 2003. Borgatti showed that centrality indicesare best understood as tightly coupled to a specific kind of network flow or networkprocess associated with it. His main idea, that centrality indices come with a modelof a network flow or process, can be generalized to other types of data mining andquality measures. I will thus discuss the question of responsibility when measuresare used in societally important algorithmic decision-making systems, such asterrorist identification systems which include social network features.