Gilles Kahn PhD prize, 2010 vintage
On 3 February 2011, in Grenoble, three young researchers will be honoured. During a joint ceremony with the French Academy of Sciences, they will officially receive the 2010 Gilles Kahn PhD prize awarded by SPECIF (the association of computer science teaching and research staff in France). The winner of the first prize, Xavier Allamigeon, has just been recruited into the MAXPLUS team at Inria Saclay - Île-de-France. The two second prize winners, Sébastien Bubeck and Stanley Durrleman, studied for their PhDs in one of the Institute's project teams before continuing post-doctoral research abroad. Spotlight on these three bright new stars from 2010.
First prize awarded to Xavier Allamigeon
The Gilles Kahn first prize has been awarded to Xavier Allamigeon
for his paper entitled «
Static analysis of memory manipulations by abstract interpretation – Tropical polyhedral algorithms,»
and their application to abstract interpretation. It was produced within the modelling and analysis laboratory for interacting systems and the EADS Innovation Works' SE/IS research team. Xavier Allamigeon was recruited in September to Inria Saclay – Ile-de-France for a three-year secondment as research scientist in the MAXPLUS project team.
«This Gilles Kahn prize primarily means recognition, I believe. When doing research, it's a bit "head down and pedal" – we hope our work has some value without necessarily having an objective view. Winning an award is all the more satisfying because it followed Inria's acceptance of my application for a secondment.
Computer science is a discipline that reaches a good compromise between theory and practice. It is possible to apply research findings by building checking software and tools, which is what I did during my PhD at EADS. That is why I wanted to continue at Inria, which for me is a benchmark institute in applied mathematics and computer sciences, and incorporates values I view as important – dynamism and open-mindedness. Inria researchers are encouraged to discover other applications for their work and open up new avenues. I see my near future in academic research, which I find very intellectually stimulating. Afterwards, as a graduate of the Corps des Mines engineering school, the decision will not depend on me alone, but I remain open to all career opportunities.» - Xavier Allamigeon .
Xavier Allamigeon's subject
A software checking tool against hacking
Bugs related to improper manipulation of software memory can have serious consequences, enabling, for example, a hacker to take over a machine. The consequences are more dramatic for more critical systems, as is the case with aircraft, rockets or nuclear power stations. Hence the significance of work such as that of Xavier Allamigeon. He has endeavoured to develop an original approach making it possible to automatically and comprehensively check for the absence of such bugs during the software testing phase. To achieve this, he established a link between subjects that seem quite separate at first sight: max-plus or tropical algebra, where addition and multiplication are replaced by operations such as "2 and 2 makes 2" or "2 times 3 is 5", and advanced automatic software checking techniques.
The key point was to demonstrate the effectiveness of tropical polyhedra to determine complex numeric properties involving use of the min and max operations on data manipulated by the software. Then, to automate this approach, an effective way to run calculations on tropical polyhedra had to be found. Xavier Allamigeon consequently offered a better understanding of their mathematical properties and designed highly-effective algorithms to run geometric operations on these objects. In line with the young researcher's wishes, this checking tool is publicly available under a freeware licence. It is intended for anyone creating software – engineers, researchers, industry practitioners, etc.
Second prize to Sébastien Bubeck
won the Gilles Kahn second prize for the promising nature of his PhD on "Bandits games and clustering foundations" (jointly supervised by Rémi Munos from the Sequel Inria research team and Cristina Butucea, a professor at Paul Painlevé Laboratory at the University of Lille 1). Two very different subjects connecting mathematics, computer science theory and statistics.
«This prize means I am recognised by the computer scientist community. I am a mathematician and, in studying for my PhD at Inria, I had a mathematician's view of an area which is at the boundary between theoretical computer science, mathematics and statistics. I enjoy being at this crossroads, finding algorithms that need to be implemented on a computer, applied to the real world. The key point to the bandit game is modelling all situations where a compromise has to be found between taking new decisions and making use of decisions known to be right. For example, in a call centre, what is the first question the agent should ask? What is the first suggestion to make to online customers? Our research is providing genuine solutions to these problems. .» - Sébastien Bubeck.
Sébastien Bubeck 's subject
Bandit games, based on the principle of action-reward, were developed in the 1950s. They model many concrete problems in applied mathematics, such as repeated decision taking in uncertain environments. Bandit algorithms are used in random phenomena that are not fully understood, such as the positioning of advertising banners on web pages. What advertising will first attract the user's attention? What would have happened if the visitor had been shown a different advertisement? These are the kind of questions that bandit game algorithms try to answer. This application has been used by companies such as Google since the 2000s. Bandit games will make possible other applications that are not yet available, such as intelligent stock management, efficient radio frequency searches or searches for communications frequencies for conversation between mobile telephones.
The second part of Sébastien Bubeck's PhD is devoted to the problem of looking for groups, or clustering, in a set of data. By defining new data categorisation methods, Sébastien Bubeck's work offers a new theoretical viewpoint on how clustering algorithms are analysed.
Second prize to Stanley Durrleman
The second Gilles Kahn prize was awarded to Stanley Durrleman for his PhD entitled "Statistical models of currents for measuring the variability of anatomical curves, surfaces and their evolution" supervised by Nicholas Ayache, Xavier Pennec and Alain Trouvé. He was a PhD student at the University of Nice - Sophia Antipolis, jointly supervised with ENS Cachan higher education and research institute, in the ASCLEPIOS Inria project team, before joining the University of Utah (USA) for his post-doctoral research.
« This prize is firstly recognition of the work I have done, even if many other PhDs are also very worthwhile. In terms of visibility, the impact is important; this prize is a boost for applications and collaboration opportunities. It provides a quality label and facilitates contacts. I'm all the more appreciative as it bears the name of Gilles Kahn. It’s symbolic because this former head of Inria worked hard on pulling together computer science and the fields of biology and healthcare. He would have appreciated the mindset with which I undertook this work with my supervisors. We have developed an essential tool for improved understanding of the variability of anatomical structures within population groups. And for understanding the impact of a disorder on these structures, highly sophisticated algorithms have to be run in order to extract and process the considerable volume of data drawn from medical images» - Stanley Durrleman
Stanley Durrleman 's subject
Better understand variety of anatomical forms seen in medical images
This PhD offers an approach to analysing and understanding the immense variety of anatomical forms seen in medical images. This makes it in particular possible to learn what separates a healthy structure from an unhealthy one. Based on methodological, numerical and algorithmic tools, this generic method – now independent of any illness or condition – makes systematic analysis of anatomical structures possible regardless of their shape (curves, surfaces, volumes, or set of points) and their topology. Another innovation resulting from Stanley Durrleman's work is statistical analysis of the change in shape from longitudinal data where each subject is observed on several occasions over time. This is, for example, one way to systematically detect autism-related developmental delays. Following this PhD, the exoShape investigative tool was made available to the scientific community.
These articles could interest you:
Created in 1998 and renamed "Gilles Kahn prize" in 2007 in tribute to the former Inria Chairman and CEO and member of the French Academy of Sciences, this is an annual prize rewarding a computer science PhD for the originality of its results, the field and methods used, the importance and impact of the findings, and the quality of writing.