Jérome Callut's homepage
 

Who am I ?

 

I obtained a Master's degree in Computer Science at the "Université Libre de Bruxelles" (ULB) in September 2003. During my studies, I went on a traineeship at the Xerox Research Center Europe (XRCE). I was supervised by Jean-Michel Renders and developed an efficient implementation of the Support Vector Machines (SVM) for classification. I'm currently a researcher funded by the FRIA at the "Département d'Ingénierie Informatique" (INGI) of the "Université Catholique de Louvain" (UCL). In this context, I'm working on a PhD thesis in the field of Machine Learning. The expected date of graduation is end of June 2007.

 

Research

 

I'm working with my advisor Pr. Pierre Dupont on the problem of modeling of a sequential process from a data sample. This kind of modeling has applications in a variety of domains such as speech recognition, text mining and biological sequence analysis to name a few. We mainly concentrate our attention on modeling dynamic features in the sequences called the First Passage Times (FPT). The FPT are fundamental characteristics of a sequential process and are closely related to the long-term dependencies possibly present in the process. We propose three novel approaches relying on the FPT to solve the following problems: (i) Hidden Markov Models (HMMs) induction (ii) sequence classification and (iii) graph mining. This work includes a tool (PHit, available soon) for fitting the FPT dynamics with discrete Phase-type distributions, a tool (K-walk) for extracting relevant subgraphs in large networks and for mining business processes, and new kernels based on the FPT and Phase-type distributions.

 

Publications

 

  • J. Callut and P. Dupont, A Markovian Approach to the Induction of Regular String Distributions, Lecture Notes in Artificial Intelligence, No. 3264, Springer-Verlag, 7th International Colloquium on Grammatical Inference (ICGI), Athens, Greece, pp. 77--90, October, 2004.

  • J. Callut and P. Dupont, Fβ support vector machines, IJCNN05, International Joint Conference on Neural Networks, Montréal, Canada, pp. 1443-1448, 2005.

  • J. Callut and P. Dupont, Séparateurs à vaste marge optimisant la fonction Fβ, CAp 2005, Conférence d'Apprentissage, Nice, Presses Universitaires de Grenoble, pp. 79-91, 2005.

  • J. Callut and P. Dupont, Inducing Hidden Markov Models to Model Long-Term Dependencies, Lecture Notes in Artificial Intelligence, No. 3720, Springer-Verlag, 16th European Conference on Machine Learning (ECML), Porto, Portugal, pp. 513-521, 2005.

  • J. Callut and P. Dupont, Learning Hidden Markov Models to Fit Long-Term Dependencies, UCL-INGI Research Report RR 2005-09, July 2005.

  • J. Callut and P. Dupont, Sequence Discrimination using Phase-type Distributions, Lecture Notes in Artificial Intelligence, No. 4212, European Conference on Machine Learning (ECML), Berlin, Germany, pp. 78-89, 2006.

  • P. Dupont, J. Callut, G. Dooms, J-N. Monette, Y. Deville, Relevant subgraph extraction from random walks in a graph, Research Report UCL/FSA/INGI RR 2006-07, November 2006.

    You can also download my Master Thesis which deals with Support Vector Machines (SVM). The document presents (in French) an implementation of the SVM that I realized during a traineeship at the Xerox Research Center Europe (XRCE).

  • J. Callut, Implémentation efficace des SVM pour la classification, Deparment of Computer Sciences, ULB, and Xerox Research Center Europe.
  •  

    Some Presentations

     

  • International Colloquium on Grammatical Inference (ICGI), Athens, Greece 2004 [PDF]
  • F.N.R.S. contact day, Brussels, Belgium, 2005 [PDF]
  • International Joint Conference on Neural Networks (IJCNN), Montreal, Canada, 2005 [PDF]
  • European Conference on Machine Learning (ECML), Berlin, Germany, 2006 [PDF]
  • Semidefinite Programming for Everyone, INGI, Belgium, 2006 [PDF]
  •  

    Teaching

     

    I'm assistant to Pr. Pierre Dupont for the INGI2262 course (Machine Learning). The reference book for the course is Tom M. Mitchell "Machine Learning" McGraw-Hill International Editions, 1997, ISBN: 0071154671.

    I'm also assistant to Pr. Philippe Delsarte and Pr. Vincent Blondel for the discrete mathematics part of the FSAB1103 course.

     

    Interests

     

    As a computer scientist, here are my points of interest :

    • Kernel methods
    • Statistical learning theory
    • Stochastic processes
    • Convex optimization
    • Hidden Markov Models (HMMs)
    • Spectral graph theory
    • Learning and mining with structured data
    • Business process modeling

    Hobbies

     

    I'm a guitarist in a jazz band called The Yellows. Our music ranges from jazz standards to funk. This young band performed a few times in jazz clubs and at a festival. We sometimes play a composition of mine called Persian Night. As a guitarist, my main influence and favourite guitar player is Philip Catherine. Other influences are Brad Mehldau (piano), John Scofield (guitar), Chet Baker (trumpet), John Coltrane (sax), Eric Clapton (vocal, guitar), Madeleine Peyroux (vocal, guitar) and Elliott Smith (vocal, guitar) to name a few.

     

    Contact

     
    Département d'Ingénierie Informatique
    Université Catholique de Louvain
    Réaumur a.337.20
    2, place Sainte Barbe
    B-1348 Louvain-la-Neuve
    tel: +32 10 47 91 16
    e-mail : jcal@info.ucl.ac.be