Professional background

Education

Research activities

Background

• Machine Learning,
• Data Mining,
• Grammatical Inference

studied objects

• Markov chains,
• Hidden Markov Models,
• Multiplicity Automata,
• Weighted Transducer,
• Tree Weighted Automata,
• Tree Edit Distance

main theoretical results

• Multiplicity Automata Algorithm which identifie the class of rational stocastic languages in the limit with probability one,
• The problem to know whether an MA generates a probabilistic distribution is undecidable (was left open in Probabilistic Automata, Azaria Paz, 1977),
• Algorithm of identification in the limit with probability one of some subclass of non deterministic HMM

Softwares

SEDiL

SEDiL: Software aimed to learn tree edit distances. Developped in JAVA (11.000 lines with javadoc comments) ;

main features:

• edition operations cost matrix learning,
• tree or string edit distance computation,
• tree and strings visualisation,
• classification using the K nearest neighbor algorithm,
• Tree pairs generation,

DEES

DEES: Software of Multiplicity Automata inference written in C++ (7.500 lines, 10.000 with comments) ;

Main features:

• Multiplicity Automata Inference,
• Probabilistic Automata Inference (Hidden Markov Model Inference ; learn Parameters and Structure),
• Deterministic Probabilistic Automata Inference
• Baum Welch Algorithm (learn parameter of an HMM from a sample),
• GraphViz export,
• Sample generation from an MA,
• And many others...

Scientific Publications

Skills

Miscellaneous

Network Project

I have done with Emmanuel Godard the conception of an anonymous network protocol (- in french -).

Mac Softwares

I've done some Mac sofwares. Such as screen saver using Quartz Composer, Dashboard widgets and Objective-C.

Web

You can see details on how I did this website.