Fabien Viger

http://www.liafa.jussieu.fr/~fabien

Interest

Education

PhD computer science: 1st semester 2007 (Doctorat d'informatique), University Pierre et Marie Curie
MS computer science, 2004 (DEA d'Algorithmique), University of Paris 6 and École Normale Supérieure
Scolarship (rank: 10) at the École Normale Supérieure d'Ulm, 2001-2005
Every year, 40 french students selected nation-wide at undergraduate level are sponsored by the Ministry of Education and have the opportunity of following high-level classes in any field they choose, during 4 years. The courses I followed are listed below.

Experience during academic research

Graphical layout of large node-centered directed networks
Example online: graph obtained with the traceroute tool from one source (big, white circle) to 400 random IP addresses. Nodes are colorized by AS.

Design and implementation (C++) of graph algorithms, during the PhD.
Typical examples: measuring the clustering, betweenness centrality, simulating traceroute sampling.. For all these algorithms, the size of the graphs (up to hundreds of millions of vertices) was a real challenge.

Efficient generation of large random connected graphs with prescribed degrees, 6-month internship during the MS.
Design and implementation (C++) of an original and simple method, based on heuristics, transforming a quadratic algorithm into log-linear, with low constants. In practice, this reduced the computational cost by more than 1000 for the graphs we needed to generate.

Extensive use of the traceroute tool
Discovery and diagnosis of most of its unexpected patterns in real-life situations. Conception of an improved traceroute, less error-prone (90% of the targeted anomalies do disappear) and made publicly available.

Active Probing with ICMP Packets, 6-month Internship at CUBIN, The University of Melbourne, supervised by Darryl Veitch, 2003.
The goal was to measure accurately the bandwidth of any targeted router on the Internet, using tricks based mainly on ICMP. The implementation was done in C and the data analysis made with Matlab. A photo of the team at that time is available here

In general: dealing with large, exotic and complex data
Typical examples are the physical Internet maps I dealt with during my PhD. On those kind of large, heterogenous networks, the only true rule is that all rules have exceptions somewhere in the network.

Selected Publications

Detection, Understanding, and Prevention of Traceroute Measurement Artifacts , at IMC 2006
Network Inference from TraceRoute Measurements: Internet Topology ‘Species’ , accepted for publication in Physical Review E
Efficient generation of random connected graphs with prescribed degrees , at COCOON 2005.

Honors

Languages