Aiello Luca Maria

title : A glimpse on social influence and link prediction in OSNs

abstract : Homophily, a property that has been observed in statical
snapshots of several social environments, is determined by complex
dynamical processes that occurs between individuals. Similarity
between actors can be induced by the information flowing along the
social pipes or, conversely, can induce the creation of ties between
people who have the same topical interests or shares the same
'confounding' features like mother tongue or home
country. Unravelling the mechanisms of social influence processes
moves from the study of the relationship between similarity and link
creation in social networks; we explored this relationship in
different feature-based Online Social Networks, where users can be
compared on the basis of the feature vectors that characterize their
public profiles. A deeper understanding of the dynamics of topical
clustering paves the way to social link prediction and to the
identification of the most informative features that determine social
aggregation.

---------------------------------------------------------------------
Aynaud Thomas

title : Static community detection algorithms for evolving networks

abstract : The study of the structure of complex networks has revealed
that they are composed of dense sub-networks called communities. Many
algorithms have been proposed to detect automatically such communities
in static networks.  However, most complex networks are evolving and
the classical approach When dealing with evolving networks is to
consider them as a succession of snapshots, each representing the
network state at a given time.
It is then possible to use well known community detection algorithms
on each snapshot. This gives a set of partitions (one for each
snapshot) and we have to reconstruct the dynamics of communities from
this set of partitions. The main problem in this context is
stability. Indeed, algorithms are often non-deterministic and a small
modification of the input network may result in major changes of the
detected community structure, making the reconstruction senseless.
To validate this reasoning we studied the stability of three well
known community detection algorithms (Walktrap, Newman's fast greedy
and the Louvain Method) using an edit distance to measure the
stability between successive partitions. This metric allows us to
conclude that none of the algorithms is stable enough to enable a
proper tracking of the communities: even after a really small change
like the removal of one random node, almost 25% of the nodes change of
community.

We have then proposed a modification of the Louvain Method to achieve a
better stability based on the initialisation of the algorithm. We shown
that this is sufficient to obtain a very good stability with only a
small loss of quality. We further modified the algorithm by introducing
a parameter that is a trade-off between stability and quality.

The behavior of the communities with this modification on an artificial
dynamic is much more satisfactory: the communities are very stable in
general but some small modifications still induce bigger (unexplained)
gaps between the partition. Finally, on real dynamics we obtain similar
results: classical algorithms are too unstable and our modification
gives more realistic results.

---------------------------------------------------------------------
Bajardi Paolo

title: Longitudinal analysis of microdynamical complex networks: a case study

abstract:Most complex networks have up to now been studied as static objects,
often for lack of longitudinal data. Many complex networks are however dynamically 
evolving entities. Case studies of empirical datasets are crucially needed in order 
to characterize the evolution of these systems, understand the effects they have 
on dynamical processes and develop new modeling frameworks. We present a case study 
of a network of the movements of Italian cattle with a daily resolution at the individual animal level. 
We first analyze the data using aggregated views on different timescales, and then focus on a 
longitudinal analysis. We show how the stationarity of statistical distributions coexist 
with dynamics at the node and link levels, on all timescales. We also show how static views 
of the displacement network hide important patterns of structural changes, with strong consequences 
for the determination of nodes' centrality or for the unfolding of dynamical processes. We finally
put forward a definition of dynamical motifs, which takes into account the causality 
of paths along the network and allows to exhibit a time arrow in the network's evolution.  


---------------------------------------------------------------------
Benamara Lamia

title: Estimating properties in dynamic systems: the case of churn in P2P networks

abstract: In many systems, such as P2P systems, the dynamicity of
participating elements, or churn, has a strong influence on the
behavior of the system.  It is therefore important to be able to
characterize it, and in particular many efforts have been made to
capture the session length distribution.  However in most cases,
estimating it rigorously is difficult.  One of the reasons is that,
because the observation window is by definition finite, we miss parts
of the sessions that begin before the window and/or end after it.
This creates a bias, which tends to decrease when the observation
window length increases.  However, it is difficult to quantify its
importance, or how fast it decreases.
We introduced a general methodology that allows us to know if the
observation window is long enough to characterize a given property.
This methodology is not particular to one study case and can be
applied to any property in a dynamic system.  We will present the
methodology and apply it to the study of session lengths in a massive
measurement of P2P activity in the eDonkey system.  We will show that
the measurement needs to last for at least one week in order to obtain
representative results.  We will also show that our methodology allows
us to precisely characterize the shape of the session length
distribution.

---------------------------------------------------------------------
Borgnat Pierre

title : A study of the Velo'V system of shared bicycles in Lyon's city

abstract : Community shared bicycle systems, such as the Velo'v
program launched in Lyon in May 2005, are public transportation
programs that can be studied as a complex system composed of
interconnected stations that exchange bicycles. They generate digital
footprints that reveal the activity in the city over time and space,
making possible a quantitative analysis of movements using bicycles in
the city. We will present a study relying on nonstationary statistical
modeling and data mining, that allows us to first model the time
evolution of the dynamics of movements with Velo'v, that is mostly
cyclostationary over the week with nonstationary evolutions over
larger time-scales, and second to disentangle the spatial patterns to
understand and visualize the flows of Velo'v bicycles in the
city. This study gives insights on the social behaviors of the users
of this intermodal transportation system, the objective being to help
in designing and planning policy in urban transportation.

---------------------------------------------------------------------
Dumas Guillaume

title : Coupled human connectomes exhibit variety of dynamical properties

abstract : With the recent development of diffusion MRI, a new
approach called connectomics has introduced a growing number of
anatomical datasets in order to enrich our understanding of brain
structure. In parallel, functional simulation in the neurodynamic
field has also enlightened our understanding of the variety of
dynamics emerging in the brain. With the combination of these two
fields the intertwined problematics of structures and dynamics can be
tackled a little further. Social neuroscience also provides a well
tuned framework for the study of coupling between complex
systems. Thanks to new hyperscanning techniques, which allow the
simultaneous recording of multiple subject while they interacts, the
measure of coupling between two EEG recordings of subjects engage in
interaction is reachable. We propose in this paper a comparison of
different functional measures of coupling. By the use of numerical
simulations we compare their behaviors and propose a new statistical c
lustering technique specifically designed for the apprehension of
these particular couplings. Application to real hyperscan data is also
provided and compared to simulations. The results show similar
dynamics and open new insights about how structural constraints could
lead to information processing specificity of brain areas.

---------------------------------------------------------------------
Gili Tommaso

title : Topological Analysis of the Default Mode Network: a Graph Theory Approach.

abstract : In typical functional MRI (fMRI) experiments a stimulus is
presented and changes in brain activity in response to the stimulus
are recorded. During performance of attention-demanding cognitive
tasks, two opposite types of responses are commonly observed. A
specific set of frontal and parietal cortical regions exhibit activity
increases whereas a different set of re- gions, including posterior
cingulate, medial and lateral parietal, and medial prefrontal cortex
Attention demanding cognitive tasks not only increase activity in
regions whose function supports task execution, but also trigger
activity decreases in regions supporting task-unrelated processes.
Although most of fMRI research deals with changes in brain activity
associated with task performance, recently the study of spontaneous
brain activity, present in the absence of a task, started being
investigated systematically. The set of brain regions that
de-activates during goal-oriented tasks has been termed "the default
mode network" (DMN). Recent studies showed correlated patterns of
slow (0.01-0.1 Hz) spontaneous fluctuations (functional connectivity)
within this network at rest, introducing to a novel neuro-anatomical
modelling of brain function. Graph theory has proved to be very useful
in describing the relationship between brain networks' nodes, and
specifically different graph models have been proposed to study the
brain's functional connectivity. The aim of this work is to explore
the properties of the DMN in the framework of graph theory and the
alterations induced on it by a cognitive task, with respect to the
resting condition.

---------------------------------------------------------------------
Grauwin Sebastian

title: Disentangling the science of complex systems through network analysis of bibliometric data

---------------------------------------------------------------------
Heymann Sebastien

title: The Gephi visualization software

abstract: During the last decade, some graph softwares tackled the challenge of
visualizing the evolution of networks in various ways. We present the
pragmatic approach implemented in Gephi, an open source software ready
for dynamic network analysis. Gephi provides built-in features for
filtering a network, given a time slice, and offer reactive user
interface through its Timeline component. One can build dynamic
statistics upon a simple API, using network structure but also nodes
and edges attributes changes over time. Finally, the graph streaming
API enables the capture of an evolving network in real time from any
datasource, and aims to synchronize and dispatch events with other
Gephi instances over the network.

---------------------------------------------------------------------
Isella Lorenzo

title: What's in a crowd? Analysis of face-to-face behavioral networks

abstract: The availability of new data sources on human mobility is opening
new avenues for investigating the interplay of social networks,
human mobility and dynamical processes such as epidemic spreading.
Here we analyze data on the time-resolved face-to-face proximity
of individuals in large-scale real-world scenarios. We compare two
settings with very different properties, a scientific conference
and a long-running museum exhibition.
We track the behavioral networks of face-to-face proximity,
and characterize them from both a static and a dynamic point of view,
exposing important differences as well as striking similarities.
We use our data to investigate the dynamics
of a susceptible-infected model for epidemic spreading that unfolds
on the dynamical networks of human proximity.
The spreading patterns are markedly different for the conference
and the museum case, and they are strongly impacted by the causal
structure of the network data. A deeper study of the spreading paths
shows that the mere knowledge of static aggregated networks
would lead to erroneous conclusions about the transmission paths
on the dynamical networks.

---------------------------------------------------------------------
Jurman Giuseppe

title : Spectral measures for biological network comparison

abstract : In a large class of complex networks, structures can evolve
along a time axis. Such evolution encodes a relevant amount of
information that in biological networks is characteristic of the
underlying biological process and crucial for applications. For
instance, comparing protein interaction networks along time is a key
step in realistic dynamic modeling of cellular functions. Network
comparison is a basic operation in modeling the dynamics of change but
its quantitative implementation is still an unsettled
problem. Quantitative strategies generally employ similarity measures
among networks as a tool for comparison. Similarity can be assessed by
local methods (e.g. searching maximal common substructures or graph
editing) or by global comparisons. Here we analyze the role of
spectral distances as a specific class of similarity measures that can
be applied to compare global topology of the underlying graph. In a
nutshell, the spectral measures are functions of the eigenvalues of
one of the possible connectivity matrices of the compared
networks. They may be combined with ad hoc kernels such as the heat
diffusion or the minimal path kernel. In this contribution, definition
and main properties of the family of spectral distances are reviewed.
Variants and applications to biological dynamical networks are then
discussed. A particular attention is directed to the integration of
spectral measures with other algebraic indicators (e.g., the dimension
of the automorphism group or other symmetry indices) to reveal major
evolutionary steps and specific structural changes along a network
time series.  (joint work with Cesare Furlanello, FBK).

---------------------------------------------------------------------
Lio Pietro

title: Investigating human cognitive heuristics by web-based communications: a chat room tool.

abstract: Human heuristics have evolved for hundred of thousand of years under
strong selective pressure. Therefore they represent higly optimized
tool for social problem solving and thus their investigation beyond
being interesting for the understanding of human mind, represent a
fundamental ingredient for cognitive based computing. The best
enviroment for studiyng heuristics is the dynamics of small group,
where everybody is able to have an individual perception of any of
others. Data extraction from actual observation of human groups is a
daunting task. We therefore study web based interactions that are
natively in digital form. In particular we developed a web instrument
that tries to furnish the digital equivalent of loud and whispering
talks, of spatial positioning into a room and some aspects of non
verbal communications. We are interested in studying non semantic data
that are supposedly more independent from cultural context. The
instrument interface is formed by two chat lines (public and private
communications), two "radars" for the private and public mental or
spatial representation of the group, and some devices for adding
emotional content and target to messages. The design of these tool has
been performed trying to accomplish the requirement of
reproducibility, availability and control of factor affectyng the
psychological dynamics. For testing such hyphothesys we performed five
experiments in controlled conditions. Each of the ten subject
composing each session, was isolated and participated anonimously, for
a duration of 45 minutes whithout any instruction (blank modality). We
measured the activity, degree of centrality and betweenness, for
eleven different dimensions (for instance: positive or negative
messagges, displacement in the radar, etc.) The different experiments
show consistent data. The main result is that the quantities under
inspection reach a stationary value after a short transient (about 10
minutes). Some quantities correlates well after an "explorative
period", as expected by psychological considerations. In particular
the exchanged messages are consistent with the mental representation
expressed by the private radar. We observed also a leadership
formation process as illustrated by the measurement of the mean
centrality degree in private radars. We are now improving the
interface in order to promote the use of nonverbal add-in, private
radars and targeted messages. Finally we are extending this instrument
by embedding task oriented tool (e.g. games, problem solving, etc) in
order to link psychological and network measurement with heuristics.

---------------------------------------------------------------------
Madan Anmol

Social Evolution: Modeling Behaviors and Opinions in Face-to-Face Networks

The exposure to new information and opinions, and their diffusion
within dynamic social networks, are important questions in education,
business, and government. However until recently there has been no
method to automatically capture fine-grained face-to-face interactions
between people and use the data to better model the adoption
process. In this talk, we describe the use of co-location and
communication sensors in 'socially aware' mobile phones to measure
and model the face-to-face interactions, opinions and behaviors of the
residents of an undergraduate dormitory for an entire academic year.

Political scientists have noted (Huckfeldt & Sprague, APSR, 1983) the
problem of mutual causation between face-to-face networks and
political opinions. For the last three months the 2008 US presidential
campaigns of Barack Obama and John McCain, automatically measured
mobile phone features can be used to measure individual exposure to
different individuals, identify political discussant ties in the
network, predict future political opinions, and capture 'dynamic
homophily' patterns, to our knowledge, hitherto unknown to political
scientists.

Similarly, face-to-face networks are the primary vehicle for
contagious diseases (Elliott, Spatial Epidemiology: Methods and
Applications, 2000), but it has not yet been possible for
epidemiologists to quantitatively measure the likelihood of contagion
as a function of contact/exposure to infected individuals in
real-world conditions (Musher, NEJM, 2003). We find that there are
characteristic changes in behaviors when individuals suffer from
common colds, influenza, stress and depression, that are reflected in
automatically captured features (e.g., total communication,
interactions relative to time of day, and movement entropy), and can
be used to estimate the likelihood of an individual being sick on a
particular day.

Finally, longitudinal studies indicate that health-related behaviors
from obesity (Christakis and Fowler, 2007) to happiness (Fowler and
Christakis, 2008) can spread through social ties. The effects of
social networks and social support on physical health are
well-documented (Berkman, 1994; Marmot and Wilkinson, 2006). We find
the mobile phone based automated exposure features predict BMI change
and obesity in this community substantially better than self-report
networks, and help understand relative contributions of exposure to
different types of individuals (i.e., individuals that are overweight,
don't exercise, or have unhealthy food habits).
---------------------------------------------------------------------
Orosz Katalin

title : Information spreading on social networks

abstract : As online communication has a great importance in our
everyday life, a huge amount of information travels and gets shared
via the internet. An item can spread if it proves to be interesting
for a given community or even for more different types of communities
which are embedded into online social networks. The large variety of
platforms and ways of being connected induce that the spreading of a
specific information can be influenced by several different factors.
We analyzed information spreading using blog and Twitter
networks. Assuming that different social networks can assist
information transmission together, we searched for relationship
between the blog network and the Twitter network. We revealed that a
Twitter and a blog network can act together promoting information
transmission. Moreover, we also show other topological quantities and
network tags connected to the spreaders with relevance to the
efficient spreading process. We studied the dynamics of the diffusion
concerning the time differences in the information flow and the
relationship between the time development and the network topology.

---------------------------------------------------------------------
Panisson Andre

title : Understanding Information Spreading on Face-to-Face Contacts
for Modeling Opportunistic/Delay-Tolerant Mobile Networks

abstract : Real world data is key to understanding how social dynamics
occurs. It is used to construct contact networks that reflect human
interactions and provide valuable information for understanding and
modeling such networks. SocioPatterns is an experimental framework to
gather data on face-to-face social interactions between
individuals. The analysis of the data gathered on different
SocioPatterns experiments can shed light on human interactions and
complex patterns of human activities, and can reveal characteristics
of the information spreading process that are stable across different
experiments and also present differences with the synthetic models
that are currently used among the community.
---------------------------------------------------------------------
Paradowski Michal B (with L. Jonak and Z. Kuscsik)

title : Tracking the diffusion of lexical innovation in online social networks

abstract : From a sociological viewpoint, linguistic creativity is a
manifestation of communities' and cultures' innovativeness. The
decisive factor in the development and spread of language is the
topology of social networks.  The uptake of novel linguistic creations
in the Internet reflects the focus of attention in contemporary public
discourse, not without its own influence on the latter (suffice it to
recollect the dynamics and main theme of status updates on Twitter
following the presidential elections in Iran, Michael Jackson's
death, Vancouver Olympic Games, and the recent Oscar gala, or the
coverage of the notorious 'Dancing Man' YouTube video in mainstream
media). The results will be presented of a data-driven study analysing
the character and speed of social diffusion of novel linguistic
formations in a microblogging site, devised with the aim of describing
the processes of the emergence of systemic order from low-level
interactions between agents inventing and imitating discrete
lexemes. Network-related innovation diffusion hypotheses
(e.g. Granovetter's (1978) threshold model of collective behaviour
and Valente's (1995) notion of network exposure) will be examined in
the context of presented case. We will demonstrate how innovation
becomes an institution, that the general innovativeness of Internet
users scales not like a power law, but a hump-shaped unimodal, and
that the exposure thresholds necessary for a user to adopt new content
from his/her neighbours concentrate at low values, suggesting that
people are more susceptible to social influence than may erstwhile
have been expected.  A dynamical systems perspective, of late
transforming many branches of science, can lead to a deeper
understanding of how mutual relations and communication between
Internet users impact the evolution of language in time and space, and
the shape and dynamics of the interactions themselves, delivering
quantitative estimates on the expansion of linguistic expressions and
the process of disambiguating their meaning, additionally allowing the
prediction of future trends and their scale.

---------------------------------------------------------------------
Perony Nicolas 

Title: Evolving communities in a real animal social network 

Abstract: The study of the structure and dynamics of interindividual
interactions is one of the major topics of social networks science,
because of their importance as the basis of complex societies. These
interactions give rise to an intricate network structure, subject to
constant evolution due to the dynamic nature of the individual
activity patterns. Of particular interest in the network are emergent
community structures, capturing tightly bound circles of individuals
sharing the same set of interests or attributes. Although the benefits
of group formation are now widely recognised, little is known of the
processes that make these communities form and dissolve, and the
influence that their dynamics have on individual fitness.  We address
this question by studying the association network of a population of
wild house mice. Each individual within the population is individually
marked with an RFID tag, to allow for their position to be recorded
whenever they enter or exit one of the 40 nestboxes present in the
barn hosting the population.  Our 2-year long dataset contains about
11 million location records for over 500 mice, accounting for more
than 1.3 million stays in all nestboxes, and leading to over 1 million
one-to-one encounters. The frequency, context and duration of these
encounters were used as a proxy for the characterisation of social
interactions.  We find that the social network of associations
consistently exhibits a highly modular pattern, which would not arise
if the associations were passive. The decomposition of the network
into communities is found to yield a number of distinct social
groups. We concentrate on the time evolution of these groups and
observe that, even though small-size communities continuously form and
dissolve within a few days, there is a number of larger-size
communities that remain in the network far longer than the typical
life cycle of an individual. We study the local structure and dynamics
of these groups, and use genetic parenting data to qualify the fitness
benefits of belonging to such durable social units. Based on the study
of the intra- and inter-community mixing patterns, we speculate on the
influence of genetic relatedness as the driving force behind the
formation of social groups, as predicted by principles of evolutionary
biology. We ask the question of whether the disintegration of a
community or its splitting into smaller units can be predicted from
the behaviour of its members through simple individual rules.  We
characterise the institutional character of long-lasting communities
in which almost all members are progressively replaced over time, but
the community persists. Finally, we speculate on the role of such
functional units as potential repositories of social knowledge,
allowing for beneficial behavioural traits to be passed throughout
generations.  This work paves the way to the integration of the
physics of complex networks with long-established principles of
behavioural sciences. Our results should be of relevance for both of
these communities.

---------------------------------------------------------------------
Seifi Massoud

title: Computing core communities in complex networks

abstract: Complex networks are generally composed of dense
sub-networks called communities. Many algorithms have been proposed to
detect automatically such communities in static networks but they are
not reliable for evolving networks due to their strongly
non-deterministic behavior.  Indeed, by running the algorithm several
times on the exact same input graph, the outcome partitions will be
generally different even if all partitions have a good quality.

We propose here to transform this non-deterministic characteristic
which is a drawback into a quality. To do so, we propose to compute
cores of communities which are sets of nodes which always belong to
the same community if a given algorithm is ran more than once on a
given network.  However, this definition is too restrictive and a
threshold x have to be used and a x-core is a set of nodes such that
all nodes are grouped together in more than x% of the experiments.

Given this definition, we studied the evolution of these x-cores over
time in simulated and real dynamics and we shown that they are very
stable and can therefore be used to describe the evolution of
sub-communities.

---------------------------------------------------------------------

Wouter Van den Broeck


Title: Visualizing SocioPatterns

Abstract:

The SocioPatterns project developed an active-RFID based sociometric platform capable of 
sensing mutual proximity in a distributed, scalable fashion. This platform enables a large-scale, 
real-time assessment of the dynamics of person-to-person interaction network with a fine temporal 
and spatial granularity. The initial focus of this project was on empirical data collection in 
support of research on dynamic social interaction networks and infection dynamics. Various experiments 
also explored the potential of the platform for more service oriented applications, and artscience contexts. The 
SocioPattens platform has been deployed in more than a dozen settings, ranging from conferences and 
trade-shows, over primary schools, long-term public exhibitions, artscience projects, and domestic settings.

We will discuss and illustrate the various visualizations developed in the context of this project, i.e.,
a real-time dynamic visualization of the localized person-to-person interaction dynamics; 
a dynamic 'social EEG'; an interactive exploration of social neighborhoods on networks that 
integrate physical person-to-person interactions with online social network data; and various 
static projections such as group flows and day-by-day temporal dynamics covering long-term deployments.

References:
- SocioPatterns project website: http://www.sociopatterns.org
- Dynamics of Person-to-Person Interactions from Distributed RFID Sensor Networks. 
Ciro Cattuto, Wouter Van den Broeck, Alain Barrat, Vittoria Colizza, Jean-Francois Pinton, Alessandro Vespignani. PLoS ONE 5(7): e11596. doi:10.1371/journal.pone.0011596, July 2010.