This position has been filled.

PhD Position

A 3-years PhD position is available at the SAMM laboratory of the Sorbonne university, in Paris. We are looking for a dynamic candidate with a master degree related to machine learning. Candidates with a strong statistical background must be fluent in R (knowledge of C/C++/Java will be appreciated). Candidates with computer science backgrounds must be fluent in C, C++ or Java. The candidate does not need to speak French if he/she is fluent in English. The student will be supervised by myself, in collaboration with assistant professors of the statistical learning and networks research axis (mainly Nicolas Bourgeois, Pierre Latouche, Madalina Olteanu and Nathalie Villa-Vialaneix).

Administrative details

after payment of health insurance, taxes, and pension, ranges from 1450€ to 1700€ depending on whether the candidate decides to give lab sessions for students
starting date
as soon as possible.
application documents
CV with references, grades obtained in the master, anything else than can demonstrate the capabilities of the applicant (master thesis for instance). Everything must be in pdf only. Applications transmitted in opaque file formats (such as word) will not be considered.

Subject: Graph data analysis with application to social sciences and history

Because graphs are simple data structures yet capable of representing complex systems, they are used in numerous scientific fields from computer science to social sciences. For instance, in biology, metabolic networks focus on representing pathways of biochemical reactions while the regulation of genes through transcriptional factors is described using regulatory networks.

Recently, there has been a growing interest in studying historical networks which are more and more easily available via digitization of various sources. Protohistorical networks can be inferred using material traces: for instance amphorae tracking is used to derive commercial exchanges between Roman or Greek antique cities, leading to a graph of those cities, decorated by the type of exchanges between them. When written sources are available, finer grained networks can be inferred, especially between persons with the so called prosopographic method. Examples of such networks studied in the SAMM laboratory include ties between ecclesiastics and nobles in the Merovingian Gaul, feudal contracts between peasants and nobles in the middle ages, exchanges between Celtic cities in the fourth century BC, etc.

Those networks lead to numerous research questions including:

  • how to compare different networks on a global point of view?
  • how to identify local differences/resemblances between different networks?
  • how to infer missing decoration on vertices and/or edges of networks?
  • how to qualify/quantify network evolution through time?
  • how to compare the evolution through time of different networks?

To address those questions, we use a large body of methods ranging from exact combinatorial algorithms to random graph models, and including non metric data mining methods. The methodological choices of the PhD will be adapted to the candidate, taking into account his/her background.

Selected SAMM articles on historical data

  • Batch kernel SOM and related Laplacian methods for social network analysis. R. Boulet, B. Jouve, F. Rossi et N. Villa. Neurocomputing, volume 71, numéro 7-9, pages 1257-1273, Mars 2008.
  • Exploration of a Large Database of French Charters with Social Network Methods. F. Rossi, N. Villa-Vialaneix et F. Hautefeuille. Dans International Medieval Congress (IMC 2011), Leeds (United Kingdom), Juillet 2011.
  • Hidden Markov models for time series of counts with excess zeros. Olteanu M., Ridgway J. Dans Proceedings of ESANN 2012 - European Symposium on Artificial Neural Networks, Belgique (2012)
  • Lexical recount between Factor Analysis and Kohonen Map: mathematical vocabulary of arithmetic in the vernacular language of the late Middle Ages, N. Bourgeois, M. Cottrel, B. Deruelle, S. Lamassé, P. Letrémy, WSOM 2012, AISC 198, pp. 255-264, 2012.
  • The random subgraph model for the analysis of an ecclesiastical network in Merovingian Gaul. Y. Jernite, P. Latouche, C. Bouveyron, P. Rivera, L. Jegou, S. Lamassé, Arxiv, (2012)
  • Cartographie de la Chronique d'Henri de Livonie, N. Bourgeois, Revue des nouvelles technologies de l'information, Janvier 2013.
  • Spatial correlation in bipartite networks. N. Villa-Vialaneix, B. Jouve, F. Rossi et F. Hautefeuille. Revue des nouvelles technologies de l'information, pages 97-110, Janvier 2013.
  • Activity Date Estimation in Timestamped Interaction Networks F. Rossi et P. Latouche. Dans Proceedings of the XXIth European Symposium on Artificial Neural Networks, Computational Intelligence and M achine Learning (ESANN 2013), pages , Bruges, Belgique, Avril 2013


2 October 2013





social sciences