Emergent patterns in nature and society

Networks: my academic muse… and some notes from Nordita conference

First a little bit of myself…

I’m not mathematician, physicist or sociologist. I don’t consider myself a network scientist. However, the topic -Networks and graph theory- appear in moments of crises, inspire me new answers and then disappear until the next cycle. Like a muse for a creative artist, although I’m far away from the art scene.

Last week I helped Örjan Bodin and Arvid Bergsten teaching a course on Networks Applications to Landscape Ecology. I thought that for the inexperienced student with little knowledge of the maths behind, networks can look a bit challenging. So, I decided to present my part on ‘Networks outside Ecology’ from an anecdotical approach, as the confused student I am rather than the expert I’m far to be.

In winter 2009 I was working on my master thesis, trying to develop minimal models or archetypes of regime shifts dynamics. After a couple of meetings with my supervisors, we realized that the work was not doable at least on the timeframe we had (still on my to do list). So, with just few months ahead and limited budget I had to make appear a new thesis proposal, make it happen and finish my master with the data already gathered, mainly causal descriptions of regime shifts.

By the time I ran into the paper Network Properties of Complex Human Disease Genes Identified through Genome-Wide Association Studies by Barrenas and colleagues. In short, they built a disease network to identify clusters of genes related with particular diseases for the human body. Genes contain the information to produce metabolites and control the whole biochemistry of your body. Hence, if one understand which clusters of genes are related with certain metabolic pathways that are more likely to cause certain diseases, then one may better understand  how to treat these diseases in a holistic way. But wait a second! What if instead of the human body I have the planet, instead of disease I have global syndromes (regime shifts), instead of metabolism I have causal processes, instead of genes I have drivers of change, and instead of treatment I have managerial options. Voilà!! my master thesis was born.

…now the interesting stuff

This simple idea have evolved into a PhD project that will keep me busy for next years. After the Resilience 2011 conference, it was clear that I have to strength my methods. I cannot identify the main global drivers of change base solely on the network topology. What if a non-central driver happens to have a strong signal? What if I focus on similar feedback loop processes to understand patterns? Is it possible to find common causal structures that leads to similar behaviors in very different systems (archetypes)? How an increase on stress in certain driver spreads on the network and increase or dampen the likelihood of cascade reactions?

I need methods, and that’s why I sneak in the conference on the Applications of Network Theory held in the Nordic Institute of Theoretical Physics, here in Stockholm. Hot topics in networks research include contagion dynamics (e.g. spread of ideas, memes, information or diseases), time-varying graphs (e.g. twitter or flight traffic networks), spatially-embedded networks (e.g. movement dynamics), community detection (e.g. modularity) and network controllability.

Two presentation were particularly inspiring for looking at causal networks and cascading effects. Albert-László Barabasi presented his recent work on controllability of complex networks. His team suggest that one can find key driving nodes by studying the network degree distribution. Their findings show that driving nodes are determined mainly by the number of incoming and outgoing links independently of where they go; and that driving nodes tend to avoid hubs. Their findings give me hope in my causal-network analysis of regime shifts. Probably, it worth keep digging on the topological issue. On the other hand, Kwang-Il Goh presented a study on how the topology of the global macroeconomic network affects the patterns of spreading of economic crises.   Their team associate the size of the economy (GDP) with links among trading partners (countries). Then, based on who your country trade with, they calculate the probability of the avalanche or the cascading effect if a partner country collapse, taking into account shared/not shared links with others. With the right data, cascading effects among regime shifts should be traceable. However, regime shifts causal networks are hierarchical (drivers are strong enough at certain scales in time and space), and that still represents a problem to be solved. I’m still searching for answers…

The archetypes idea got revitalized with the talks by Martin Rosvall and Sebastian  Bernhardsson. Rosvall and Bergstrom used a link-based community detection technic performing random walks on an academic citation network. Their work manage to bring a messy dataset of person to person collaboration to clear patterns of academic fields interactions. Interestingly, their network is hierarchical and they approach the problem of aggregation by using a coding technique. I’m not quite sure of the usefulness of random walks on regime shifts networks. But I’d expect similar feedback processes to get cluster as disciplines get cluster in Rosvall’s network. On the other hand, Bernhardsson and colleagues study the evolution of metabolic networks by using the concept of ‘Organism Degree’ OD -shared common structure in metabolic networks among different species. The equivalent ‘Regime Shift Degree’ will link archetype structures or causality patterns in different regime shifts networks.

To close a bit of inspiration for you:

Networks has been for me a great source of inspiration in the last couple of years. For that reason, my take-home message for the network course was: get inspired and go for it. So, here is a couple of hints for you. This video published by Wall Street Journal a couple of weeks ago summarizes how networks has been recently used to study human behaviour using your mobile phone. Yes, they are watching us..

The network figure above shows how ingredients has been commonly associated in different world cuisines (high-res pdf here). As you can see, beer, coffee and tea are central nodes with high betweenness since they are common used no matter where are you reading this. You might probably have a closer look next time you leave your creativity take off in the kitchen. It might worth to double check which ingredients go better with which ones. Happy cooking!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s