# Dr. Maximilien Danisch

PostDoc at Telecom ParisTech / INFRES / IC2

Office: C212, 45 rue Barrault, 75013 Paris.
Tel: +33 6 03 42 40 01
Email: maximilien.danisch@gmail.com

RESEARCH

PUBLICATIONS

SPECIALS

Since september 2015, I have been a PostDoc at Telecom ParisTech under the supervision of Mauro Sozio. Before that I was a Ph.D candidate in computer science at complexnetworks.fr under the supervision of Jean-Loup Guillaume and Benedicte Le Grand. In 2010, I obtained a master's degree in physics from ENS Cachan after an internship at the City College of New York under the supervision of Hernan A. Makse. In 2010-2011 I spent one year as an intern at Columbia University under the supervision of Tony Jebara.

My research interests revolve around the study of very large real-world graphs.
Curriculum vitae: pdf.
PhD manuscript (in french): pdf, slides (in english): pdf.

## RESEARCH

### Softwares I've developed:

My github profile is here.
• Efficient C code:
• Parallel C code for enumerating all k-cliques in a graph. The program scales to real-world networks containing several billions of edges.
Friendster has exactly 487,090,833,092,739 10-cliques, that is a bit less than 0.5 quadrillions 10-cliques.
Also contains C code for computing the k-clique core decomposition of a graph as well as a k-approximation of the k-clique densest subgraph. The program scales to real-world networks containing several billions of edges.
• A C code for computing a Heaviest k-Subgraph (that is a subgraph containing k nodes such that the sum of the weight on its edges is maximized) in a weighted graph. The program scales to real-world networks containing several billions of edges and for k up to 10, 20 or more depending on the structure of the graph. Approximated version of the problem can be solved for larger k.
A stackexchange question on the topic.
• A C code to compute all overlapping communities in a network following a from local to global approach. The program scales to networks containing several millions of nodes and several hundreds of millions of edges.
• A C code to evaluate the similarity between two overlapping community structures. The program scales to billions of communities.
• Application for measuring influence on Twitter taking into account social capitalism: DDPapp.
• Application for mention recommation in order to better propagate a tweet: EasyMention.

### Softwares I find useful:

• The Louvain Method implemented in C++ by Jean-Loup Guillaume and Etienne Lefbvre as detailed in Arxiv.
• NetworkX, a Python language software package for the creation, manipulation, and study of the structure, dynamics, and functions of complex networks.
• sklearn, a Python library for Machine Learning.
• Commercial LP, QP and SDP solvers (free for academics): MOSEK and GUROBI.
• CVXPY, a python-embedded modeling language for convex optimization problems. It can be used with commercial solvers or free slovers like CVXOPT.
• I use OpenMP to make my C programs parallel by only adding a few lines of codes.

### Datasets:

• Two collections of large networks: konect and snap.
• Machine Learning datasets repository: UCI.

### Projects:

• The CODDDE project aims at better understanding the evolution of real-world complex networks. Three topics are studied within this project: community structure evolution, spreading of information and detection of unexpected changes in their structure.

## PUBLICATIONS

My DBLP profile is here.
here.

### International Journal:

1. Multi-ego-centered communities in practice.
M. Danisch, J.-L. Guillaume, B. Le Grand.
SNAM 2013.
Team website.
2. Towards multi-ego-centered communities: a node similarity approach.
M. Danisch, J.-L. Guillaume and B. Le Grand.
Int. J. Web Based Communities, Vol. 9, No. 3, pp.299-322, (2013). Team website.
3. Calculation of the Voronoi boundary for lens-shaped particles and spherocylinders.
L. Portal, M. Danisch, A. Baule, R. Mari and H. A. Makse.
J. Stat. Mech. (2013) P11009. Arxiv.
4. Model of random packings of different size balls.
M. Danisch, Y. Jin, and H. A. Makse.
Physical Review E 81.5 (2010): 051303. Arxiv.
5. ### International Conference:

6. Large Scale Density-friendly Graph Decomposition via Convex Programming.
Maximilien Danisch, T-H. Hubert Chan and Mauro Sozio.
WWW 2017. PDF.
7. On the Role of Mentions on Tweet Virality.
Soumajit Pramanik, Qinna Wang, Maximilien Danisch, Anand Kumar, Sumanth Bandi, Jean-Loup Guillaume and Bivas Mitra.
DSAA 2016.
8. Local triangle-densest subgraphs.
Raman Samusevich, Maximilien Danisch and Mauro Sozio.
ASONAM 2016.
9. A reliable and evolutive web application to detect social capitalists in Twitter.
Nicolas Dugué, Anthony Perez, Maximilien Danisch, Florian Bridoux, Amélie Daviau, Tennessy Kolubako, Simon Munier and Hugo Durbano.
ASONAM 2015 (Demo track). Team website.
Dedicated website: DDPapp.
10. Learning a Proximity Measure to Complete a Community.
M. Danisch, J.-L. Guillaume and B. Le Grand.
DSAA 2014. Team website.
11. On the importance of considering social capitalism when measuring influence on Twitter.
M. Danisch, N. Dugué and A. Perez.
BESC 2014. Team website.
12. Direct Generation of Random Graphs Exactly Realising a Prescribed Degree Sequence.
CASoN 2014. Team website.
13. Mining bipartite graphs to improve semantic pedophile activity detection.
R. Fournier and M. Danisch.
RCIS 2014 (short paper). Team website.
14. ### International Workshop:

15. Finding Heaviest k-Subgraphs and Events in Social Media.
Matthaios Letsios, Oana Balalau, Maximilien Danisch, Emmanuel Orsini and Mauro Sozio.
DaMNet 2016.
16. Unfolding ego-centered community structures with "a similarity approach".
M. Danisch, J.-L. Guillaume and B. Le Grand.
Complex Networks IV. Springer Berlin Heidelberg, 2013. 145-153. Team website.
17. ### International Book Chapter:

18. Multi-ego-centered communities.
M. Danisch, J.-L. Guillaume and B. Le Grand.
ComNetBook. Team website.
19. ### National Conference:

20. Déplier la structure communautaire d'un réseau en mesurant la proximité aux représentants de communauté
M. Danisch, J.-L. Guillaume and B. Le Grand.
MARAMI 2015--Modèles et Analyse des Réseaux : Approches Mathématiques et Informatiques. Team website
C codes: proxallcoms.tar.gz, compare.tar.gz.
21. Augmenter les retweets sur Twitter : comment tirer parti des mentions ?
S. Pramanik, Q. Wang, M. Danisch, M. Sharma, S. Bandi, J.-L. Guillaume, S. Raux, B. Mitra.
MARAMI 2015--Modèles et Analyse des Réseaux : Approches Mathématiques et Informatiques. Team website.
22. Structures biparties et communautés recouvrantes des graphes de terrains.
R. Tackx, M. Danisch and F. Tarissan.
MARAMI 2014--Modèles et Analyse des Réseaux : Approches Mathématiques et Informatiques. Team website.
23. Prendre en compte le capitalisme social dans la mesure de l'influence sur Twitter.
M. Danisch, N. Dugué and A. Perez.
MARAMI 2014--Modèles et Analyse des Réseaux : Approches Mathématiques et Informatiques. Team website.
24. Complétion de communautés par l'apprentissage d'une mesure de proximité.
M. Danisch, J.-L. Guillaume and B. Le Grand.
Proceedings of ALGOTEL'14 (16emes Rencontres Francophones sur les Aspects Algorithmiques de Télécommunications), 2014, Ile de Ré, France. HAL.
25. Une approche à base de similarité pour la détection de communautés egocentrées.
M. Danisch, J.-L. Guillaume and B. Le Grand.
Proceedings of ALGOTEL'13 (15emes Rencontres Francophones sur les Aspects Algorithmiques de Télécommunications), 2013, Pornic, France. HAL.
26. ### National Workshop:

27. Déplier les structures communautaires egocentrées - une approche à base de similarité.
M. Danisch, J.-L. Guillaume and B. Le Grand.
Atelier Fouille de Grands Graphes, EGC 2013.

## SPECIALS

• Free online classes, mostly computer-science/science, but not only.
So far I've followed:
• Machine Learning, by Andrew Ng.
• Introduction to Artificial Intelligence, by Sebastian Thrun and Peter Norvig.
• Model Thinking, by Scott E. Page.
• Game Theory, by Matthew O. Jackson and Yoav Shoham.
• Algorithms: Design and Analysis, Part 1, by Tim Roughgarden.
• Algorithms: Design and Analysis, Part 2, by Tim Roughgarden.
I enjoyed all of them and highly recommend them.
• Machine Learning and Data Science competitions: kaggle and datascience.