Antoine Cornuéjols

Geo Visitors Map
  Professeur / Full Professor

Dept. MMIP (Modélisation Mathématique, Informatique et Physique)
16, rue Claude Bernard
75005 Paris Cedex (France)

Tel     :  +33 (0)1-44-08-72-29
email   :  antoine.cornuejols [at]


Courses (in French) Links Research contracts Culture


and traineeships


I am Full Professor in Computer Science at AgroParisTech (INA-PG), associated with the "Modélisation Mathématique, Informatique et Physique" (MMIP) research department, and a corresponding member of the TAO research team at LRI (Laboratoire de Recherche en Informatique) at the Université de Paris-Sud, Orsay (France). My fields of research lie in Machine Learning, data mining and Cognitive Science. I am specially interested in the following topics :


Fundamentals of learning  (more theoretically oriented)


The geometry of learning

  • What is the space of programs when the change operator is learning from examples ?
    •  What kind of dot products can yield negative values (as opposed to the mutual information which is not fit for learning) ?
    •  What is a proper metrics ?
    •  How does order sensitivity in learning relate to curved space ?
  • Teachability
    •  How to best order a data sequence for a specific goal ?
    •  Which data should be presented ?
    •  In which order ?
    •  At which speed ?
  • On-line learning
    •  Incremental learning (in a stationary environment)
    •  With covariate shift
    •  Tracking
    •  Concept-drift

Transfer of information

  • Analogical reasoning
    • Which foundations for analogical reasoning ?
    • Which links with inductive reasoning ?
  • Learning of new conceptual domains
    • Which articulations with known domains allow to learn a new conceptual domain ?
    • How a conceptual domain in the learning come to get integrated with other conceptual domains ?
  • What are the effects of limited transmission bandwith in a collection of learning agents ?



Phase transition in inductive learning

  • What is the influence of language on learning new knowledge ?
  • How is inductive supervised learning affected by the description language ?
    • What explanation for phase transitions in supervised learning ?
    • How to alleviate these with changes of representation language ?


Application oriented methodological developments


Applications in data mining

  •  Scene analysis in changing environments
    •  Control of the memory in on-line learning
    •  Tracking
  •  Contour detection in the dynamical recognition of a moving kidney
    •  Boosting methods
  •  Text analysisin a risk assessment task
    •  Automatic detection of parts of texts related to a risk assessement qualification
    •  Automatic assessement of uncertainty in parts of text
  •  Protein docking
    •  Learning from positive instances only
    •  Reduction of the false positive rate
    •  Use of abstraction and change of granularity in data mining methods
  • Study of genomics and microarray data
    •  Combination of weak selection methods
    •  Active methods for feature selection
  •  Study of cardiovascular accidents (prediction of)
    •  Methods for learning in the presence of imbalanced data sets and asymetrical labelling noise
    •  Asymetrical semi-supervised learning


Courses (in French) Links Research contracts Culture