(Partial) conference recap: Applications of Network Theory, Stockholm 110407-09
I’ve spent a couple of days at the Conference on Applications of Network Theory, which was held in Stockholm. Since I live here anyway and had some domestic matters to attend to, I was only able to attend the first two days (and missed several lectures on the second day), but I still enjoyed the conference a lot. As my co-blogger Joel put it, the ideal conference would be a mash-up between this conference and Strata (although my ideal conference would also have a dedicated life science big data track). Some tweets were tweeted from this conference using the #norditanwk hashtag.
The name of the conference describes the general contents quite well. Participants described the application of network theory to a wide variety of things, from Yong-Yeul Ahn’s flavor network, which connects cooking ingredients based on the number of chemical flavor compounds they have in common, via Kwang-Il Goh’s simulations of global trade networks and how economic troubles spread in them based on their topology (see e g Impact of the Topology of Global Macroeconomic Network on the Spreading of Economic Crises) to Örjan Bodin‘s ecological/social networks that model systems like small-scale fisheries in east Africa.
Many talks dealt with the problem of community detection in networks. This is not a simple matter – Aaron Clauset remarked that he thought there is at least 50 years or so left until this problem has been fully addressed. Sune Lehmann said that networks with “pervasive overlap”, where nodes (e g people) belong to many groups, are very hard to decompose into meaningful communities. His solution (see Link communities reveal multiscale complexity in networks) is to look at communities on the level of links (relationships) instead of nodes.
Other talks explored various kinds of refinements to standard network models. Vincent Blondel (who has developed the popular Louvain community detection method) talked about overlaying communities onto geographical maps, and how to cancel out geographical effects in communities. Jari Saramäki talked about the often-overlooked temporal nature of many networks, where nodes are not constantly connected but rather interact via transient events. He has tried to generalize notions such as network centrality to temporal graphs.
Renaud Lambiotte talked about how he analyzed data from an online game (pardus.at) where all player actions are stored in log files. Interaction data between players (in the form of messages, trading events, attacks etc.) were used to characterize positive and antagonistic relationships, and it turned out that these were not mirror images of each other but rather had quite different properties. Lambiotte has also analyzed the personality of popular Facebook users and found that extroversion and age are the only significant predictors of popularity (high age is a negative predictor.)
Veronica Ramenzoni talked about interesting experiments where she has studied how people get “in sync” with each other through mirroring and other effects, and Fredrik Liljeros discussed human sexual networks and venereal disease transmission, claiming that Casanova-type sexual behaviour may be better than polygamy when it comes to the spreading of HIV. (That is, Casanova behaviour would lead to less spreading.) Helena Buhr talked about network formation among future business elites, studied through data on gifts given by students to top-ranked MBA schools.
Apologies to any speaker who I didn’t mention here; as mentioned above, I missed the whole third day and the end of the second day. It was fun to catch up with what’s happening at the frontier of network science!