Postdoctoral Position in Machine Learning for Human-Robot Interaction at University of Lille, France
TopicMachine Learning for Human-Robot Interaction
Job description
This position is offered in the framework of the BabyRobot H2020 (www.babyrobot.eu) project involving 8 partners across Europe (including France, Sweden, UK, Greece, Germany and Denmark). The project focuses on interactive robotics and especially on interaction with children.
Breakthroughs in core robotic technologies are needed to support this research mainly in the areas of motion planning and control in constrained spaces, gestural kinematics, sensorimotor learning and adaptation. In addition, new models of interaction need to be developed. Because of the human being in the loop, standard control theory can hardly be applied. For this reason, machine learning methods such as reinforcement and imitation learning have been identified as candidates to address these issues in a unifed framework. Therefore, the applicant will be involved in research in core machine learning applied to control and interaction. Several directions of research are envisioned. First, recent works on stochastic games [3] applied to dialogue management [1] can be further investigated so as to be adapted to multimodal and multiparty interaction scenarios. Second, the "learning from demonstration" (LfD) [4] paradigm can be adapted to the adversarial case so as to transfer interactional behaviours from actual human-human interactions to machines. Other topics can be investigated such as inverse reinforcement learning [2] or transfer learning.
ProfileThe applicant should have completed a PhD in computer science, statistical learning or robotics. The ideal candidate will have a strong background in machine learning and especially in reinforcement learning or stochastic games. Experience in interactive systems (spoken dialogue systems, interactive robotics, human-machine interfaces) would be much appreciated. The recruited person will be involved in the management of the project, participate to consortium meetings and contribute to deliverables. Therefore, good communication skills and autonomy are mandatory. Preference will go to candidate with a strong publication record.
Work environmentThe position is offered in the Sequential Learning (SequeL) research team (joint team between Inria, Univ. Lille and CNRS) located in Lille, France. SequeL is a world-leading group in reinforcement learning, bandit theory and recommendation systems involving 30 members (including 10 permanent staff members). The team's working language is English. The team is part of the French National Institute for Computer Science and Mathematics (Inria) as well as the Computer Science and Signal Processing laboratory of Lille (CRIStAL). Lille is the capital of the north of France, a metropolis with 1 million inhabitants, with excellent train connection to Brussels (30 min), Paris (1h) and London (1h30).
How to applyThe application should include a brief description of research interests and past experience, a CV, degrees and grades, motivation letter, relevant publications, letter(s) of recommendation and contact information to reference persons.
Details
Application deadline: 30th of April 2016
Starting date: May or June 2016
Duration: 24 months (can be extended)
Salary (after taxes): 2100 euros
ContactsOlivier Pietquin: olivier.pietquin@univ-lille1.
Bilal Piot: bilal.piot@univ-lille1.fr
References[1] Merwan Barlier, Julien Perolat, Romain Laroche, and Olivier Pietquin. Human-machine dialogue as a stochastic game. In Proceedings of the 16th Annual SIGdial Meeting on Discourse and Dialogue (SIGDIAL 2015), pages 2{11, Prague (Czech Republic), September 2015.
[2] Edouard Klein, Matthieu Geist, Bilal PIOT, and Olivier Pietquin. Inverse Reinforcement Learning through Structured Classification. In Advances in Neural Information Processing Systems (NIPS 2012), pages 1007{1015, Lake Tahoe (NV, USA), December 2012.
[3] Julien Perolat, Bruno Scherrer, Bilal Piot, and Olivier Pietquin. Approximate dynamic programming for two-player zero-sum markov games. In Proceedings of the International Conference on Machine Learning (ICML 2015), Lille (France), July 2015.
[4] Bilal Piot, Matthieu Geist, and Olivier Pietquin. Learning from demonstrations: Is it worth estimating a reward function? In Hendrik Blockeel, Kristian Kersting, Siegfried Nijssen, and Filip Zelezny, editors, Proceedings of the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML/PKDD 2013), volume 8188 of Lecture Notes in Computer Science, pages 17{32, Prague (Czech Republic), September 2013. Springer.