Gérard DREYFUS, professor, ESPCI-ParisTech
Home page
Gérard Dreyfus had two scientific lives in succession: trained as a physicist, he investigates the charge storage properties of insulating polymers (electrets); he contributes to the development of various applications of electrets (microphones, ionizing radiation detectors, electrostatic relays), and to the theoretical and experimental development of a non-destructive method for charge density measurements in insulators, based on the propagation of a shock wave in the material. The method is widely used and still undergoes various developments.
Hired as full professor in 1982, he initiates a completely new research activity and creates a group devoted to machine learning and to the neural bases of information coding in the brain.
As a pioneer of machine learning, he publishes with his collaborators, as early as in 1985, a series of fundamental articles on the properties and training of Hopfield networks. In 1988, he organizes the first European conference on neural networks nEuro'88 in Paris, France. Over the years, with his collaborators and doctoral students, he publishes a number of basic results, essentially methodological in nature: representation and training of dynamic models, variable selection, model selection, semi-physical modeling, learning from structured data, etc. The methods and results are applied to a large number of industrial, financial and medical applications, developed in the framework of collaborations with companies or government institutions, French and international.
Author or co-author of over 200 publications in international journals and conferences, he is co-holder of more than 20 French and international patents. Motivated by the dissemination of knowledge, he is co-author of two textbooks in French on machine learning (1, 2), and of a textbook in English published by Springer. He is associated to the direction of the collection "Algorithmes" of the French publisher Eyrolles.
He is co-founder of the machine learning software publisher Netral S.A., in existence since 1994. Three of his PhD students created companies that are active in machine learning: Stefan Knerr (Vision Objects), Isabelle Guyon (Clopinet) and Rémi Dubois (Ciprian).
Gérard Dreyfus plays a role in the machine learning community at the international level. In 2003, he creates the French chapter of the IEEE Computational Intelligence Society. He sits on various technical committees of the IEEE. He is a member of the organizing and/or program committees of several international conferences, including the prestigious "Learning Workshop" from 1988 to 2008. He is associate editor of several international journals (Neural Computation, IEEE Transactions on Neural Networks, Neurocomputing, Applied Intelligence, Theoretical Computer Science, ...).