logobas
Complex Networks team
http://complexnetworks.fr
LIP6 laboratory (CNRS, Universit´e Pierre et Marie Curie)
logobas
Networks/graphs from different contexts
computer science: internet, P2P, web, usages, etc.
social sciences: friendships, communications, collaborations, exchanges, economics, etc.
biology: brain, genes, proteins, ecosystems, etc.
linguistics: synonymy, co-occurrences, etc.
transportation: roads, air, electrical networks, etc etc
Various contexts, but common properties
common problems to solve
logobas
Some common questions
Measurement and metrology
How to acquire data about these networks? Reliability?
Analysis
How to describe the structure of very large networks?
Modelling
Generate artificial networks resembling a given network
⇒Goals: understanding, simulations, . . . Algorithmic questions
Efficient computations on very large networks
logobas
Dynamics – 1
(C.Magnien, M.Latapy)
Generic dynamics
How todescribethe dynamics?
evolution speed?
variety of behaviours?
. . .
Algorithmic questions How to define and compute:
degree, clique
distance (shortest path) connectedness, spanning trees . . .
Define equivalent notions to what exists in graph theory
logobas
Dynamics – 2 (M.Latapy)
Many dynamic networks are a sequence of interactions
a b c d
0 5 10 15 20 temps
e
An equivalent to the notion of community Find relevant sub-streams:
group of links which are dense both structurally and in time ex: thread in a forum, conversation in a mailing list
logobas
Dynamics – 3 & 4 (M.Latapy)
Event detection
Define a notion of normal bahviour in a stream Find deviation to normality
Find specific eventsin a graph dynamics IP traffic
Study IP traffic at a very large scale:
million active links per minute detect abnormalities, faults, attacks. . . Bitcoins
Study financial transactions at a very large scale:
100M transactions available (100K/day)
detect frauds, misappropriations, stock exchange events. . .
logobas
Towards more realistic bipartite models (F. Tarissan)
Random generation :
Erdös Rényi
Con g.
Model size and
density
+ degree
distribution +
Random Bipartite local density +
Real graph redundancy+ ?
A B C D
α β
3 2
1 4 5 6 t
A B C D
α β
3 2
1 4 5 6
A B C D
3 2
1 4 5 6 t
Rand Bip A B C D
5 2
1 3 4 6
Questions :Whichtripartiterandomization?
Theoretical issues: Expected properties?
Algorithmic issues: How to enumerate efficiently the overlapping patterns?
logobas
Link prediction and centrality distances (L.Tabourier, G.Tredan)
Use structural information to predict link appearances/disappearances
Quality evaluation: does a link actually appear or not?
⇒ measuring the graph edit distance
Other notions of distance, based on centralities
logobas
Multiscale Analysis of Dynamical Networks (R. Lamarche-Perrin, M. Latapy)
Going from the aggregation of static graphs...
→
...to the aggregation of dynamic graphs
a bc d
0 5 10 15 20 time
e
Spacial aggregate Temporal aggregate
logobas
Opinion Dynamics & Political Conflicts in the Media (R. Lamarche-Perrin, M. Latapy)
Opinion dynamics in social media (Twitter, Facebook, Instagram) Analysis of interaction patterns:
polarisation effects, echo chambers, filter bubbles, information leadership
Conflicting world views inmass media(printed press)
Study the co-occurrences of countries in articles from different newspapers
logobas
Conclusion
All topics require
some data manipulation some formal approaches
taste in interdisciplinary matters To be discussed with the candidate All topics can lead to a PhD
More details: http://www.complexnetworks.fr/projects/
Contact: [email protected]
