Lab: Imagine, Université Paris Est, France
Contact: Jean-Yves Audibert (audibert at imagine.enpc.fr)
Numerous problems linked to the exploration-exploitation trade-off remain open (even in the well-known *multi-armed bandit* problem). We want to address them, i.e. propose algorithms solving them and study their theoretical properties. In spirit, this is related to (i) the recent works on policies based on upper confidence bounds and (ii) the nice book: Prediction, learning, and games by Nicolò Cesa-Bianchi and Gábor Lugosi. The work is mainly theoretical but, as far as possible, we also intend to address some of the energy minimization problems that arise in Computer Vision, in particular combinatorial problems of 3D reconstruction, by using the proposed algorithms.
The position requires strong theoretical skills in probability and statistics. An ability to implement algorithms working on real-world data is also desirable.
To apply, please email CV, names and email addresses of 2-3
referees to Jean-Yves Audibert (audibert at imagine.enpc.fr) before May 15th, 2010.
The position is funded by a grant from the French National Research Agency (ANR). The successful candidate will have no mandatory teaching or administrative duties.