A howto on a sysadmin's dream: how to set up a KDE 2 Desktop so that your users can't change them (e.g. enforce a common color scheme, background picture, Desktop icons set). KDE 3 introduced 'Kiosk' which made this Howto obsolete.

more >>

A talk I gave on the open-source software infrastructure I set up in a primary school, in Strasbourg.

more >>

My internship report, summarizing the work done during my 6-month stay at the Swartz Center for Computational Neuroscience (University of California, San Diego).

This report summarizes preliminary results from a pilot experiment to study the interaction of proprioceptive and visual input in achieving precise arm movements. The goal of the present work is to prepare future experiments with Parkinson's disease patients that will help to better understand the role of basal ganglia in visual and proprioceptive input interaction. We analyze EEG recording of a healthy subject performing a reaching experiment with no visual input. We first investigate what kind of biases lead the subject to miss the target. Using Independent Component Analysis (ICA), time-frequency analysis and time-series warping, we identify preliminary movement-related dynamics. A potential inversion preceding movement was uncovered, and so were components with a different spectral behavior with respect to the target location and the original arm position.

more >>

Chess tournaments are competitions that involve dozens to thousands of players. Every participant plays a fixed number of games (typically 7 to 9), and the outcome of the game determines the next opponent, where winning a game makes you play against a stronger opponent while loosing a game lets you meet a weaker player on the next round. Given that scheme, a well-known tournament strategy consists in loosing the first game on purpose in order to play against weaker opponents and have an easier tournament, and only catch up with the leader towards the end of the tournament to beat them after they have been struggling to stay on top during most of the tournament. But in practice, a very small minority of players actually follow this strategy, which tends to be considered too 'risky'.

Here, I used techniques from multivariate exploratory statistical analysis such as the principal component analysis in order to test whether the strategy that consists in loosing the first game on purpose was actually productive in a real chess tournament. In the two tournaments I examined, the outcome of all but the first game correlated with the player ranking at the end of the tournament. This result is consistent with the idea that the strategy that consists in loosing the first game on purpose may be viable in practice.

more >>

My master thesis, under the supervision of Korbinian Strimmer at Ludwig Maximilian Universität in Munich, Germany. It explores how entropy and other information theoretic quantities may be used to reverse-engineer genetic regulatory networks from repeated microarray data. The problem of differentiating genes that undergo direct co-regulation from genes whose expression is similar because they belong to the same regulatory pathway is studied from a graphical modeling viewpoint. This leads to the criteria of conditional independence which can be evaluated by computing the conditional mutual information. The latter is completely characterized by the sum of the entropies of joint variables, underlining the need for an entropy estimator that is accurate even in low sampling conditions.

We introduce a new plug-in entropy estimator obtained from shrinking maximum likelihood multinomial proportions estimates to the maximum entropy target. We derive the closely related ZIPshrink and ZINBshrink entropy estimators which enhance the shrinkage estimator by first adjusting the shrinkage target depending on the fraction of structural zeros in the multinomial model. The fraction of structural zeros is estimated using a Zero-Inflated Poisson or Zero-Inflated Negative Binomial distribution to model the histogram of bin counts. We compare these three new estimators to state of the art estimators. We show that they give acceptable estimates even in the low sampling regime and are as accurate as the best estimator available today while being 100 faster, making it more suitable for large scale computations. We then compare existing approximations of conditional independence networks such as 0-1 networks and a data processing inequality based approach. As a conclusion, we briefly consider limitations of the method as well as issues related to unobserved variables, causal inference and time series as opposed to steady state experiments.

Part I serves both as an introduction and a motivation. It presents the notions of conditional independence and explains why entropy estimation is critical to genetic regulatory network inference. Part II has the core results of this report: it reviews existing entropy estimators for the discrete case, introduces a new entropy estimator based on the statistical notion of shrinkage and compares their performance. Finally, part III compares data processing inequality based approach to genetic regulatory networks reverse-engineering with the so-called 0-1 networks approach. It also has considerations about limitations, pitfalls and possible extensions of the method.

more >>

A report of what I learned while spending a week at the Observatoire de Haute-Provence (southern France), in August 2001. Thanks to Agnès Acker for the fascinating time!

more >>

An open-source PHP-based (web)link directory management system. I started that project in 2001 and was then joined by different programmers world-wide who translated it to 6 languages. I stopped maintaining it two years later. At that point, it was used in several intranets, in the university of Karlsruhe Vikar project, and on a number of other websites. Many thanks to David "dbu" Buchmann who kindly took over the project.

more >>

The report describes the QueryBuilder part of the SBDB project, for Bayer Schering. Its purpose is to generate SQL queries from criteria and patterns it receives from the user interface without knowing the database it should work on in advance. In other words, the table layout is not hard coded in the QueryBuilder but gets loaded from an external XML file as the QueryBuilder starts. In the case of SBDB, there were over 180+ tables, which made such automated graph-theoretic approaches necessary. Furthermore, the QueryBuilder was designed to run on relational database schema that are derived from object models and thus supports heritage relationships between tables.

I wrote the QueryBuilder in Java during a one-month internship at the Structural Biology and Genomics Laboratory (IGBMC Strasbourg), under the supervision of Olivier Camus and Jean-Marie Wurtz.

more >>

A Java applet that simulates nuclear magnetic resonance (NMR) of a 1D object. Third year engineering school project together with Marianne Crevon, Patrice Fayolle and Karoll Frindel.

more >>

The slides of a talk on the implementation of system calls in the Linux kernel I gave while taking the "Linux Internals" Proseminar at the CS department of Universität Karlsruhe, in the Winter 2002 / 2003.

more >>

An agent-based simulation of a planet where each pole is home to an ant colony. Ants collectively gather food and bring it back to their home colony using pheromones. The simulation was implemented for an C++/OpenGL programing class assignment.

more >>

A report I wrote for a humanities project while studying for my masters, under the supervision of Michel Faucheux. Largely based on Durkheim's sociology, we start by highlighting the social aspects underlying the phenomenon of belief. Then, we show that beliefs are also the object of science and sociological/anthropological knowledge, and that it is even a building block of knowledge and science. Finally, we conclude that beliefs mediate our knowledge and our relationship to truth.

more >>