Postdoc: Sequential Budgeted Learning (expired)

The position is closed


We are looking for an excellent Postdoc to work on the development of new sequential machine learning methods for budgeted learning in the context of large scale datasets and robotics.

In a world where the amount of produced data is growing exponentially, conventional Machine Learning (ML) models should be rethought. They have been developed on a centralized paradigm where information can be stored on a single computer (or a cluster of computers) and can be processed as a whole. Since, ML models have been adapted to the increase of the amount of information by using techniques like parallelization or sampling that do not directly integrate constraints like CPU or memory consumption. Indeed, because the quantity of produced information is growing, and because the way we access information is changing, there is right now a huge need to propose learning approaches able to integrate during learning and inference processes both, how information has to be acquired and also realistic budgeted constraints. The post-doc will focus on the development of sequential learning models able to simultaneously learn which information to acquire for any particular task, which representation to build upon the acquired information, and to compute a prediction or to learn a particular model depending if our sequential model is used ''during training'' or ''during inference'' under a set of budgeted constraints. Applications to recommender systems and robotics will be developed.
The position is funded by a national research project (Onbul Project for one year + one possible additional year.

The successful candidate will join the machine learning and information access team of the University Pierre et Marie Curie, Paris, France, led by Prof. Patrick Gallinari, and will also be associated with the Robotics Institute. The post-doc will be supervised by Prof. Ludovic Denoyer. Our research explores a number of different issues such as: representation learning for relational data, for dynamics data, reinforcement learning, deep neural networks. For a more detailed description the interested candidates may take a look at: and the list of publications within there. The University Pierre et Marie Curie is referenced as the first French university in the academic ranking of international universities. It offers considerable opportunities for training and exposure to data mining and machine learning, with a number of research teams being active on these and related fields. In addition the selected candidate will have ample opportunities to participate in the main ML conferences.
Related publications (in the team):

G. Contardo, L. Denoyer, T. Artieres, and P. Gallinari, "Learning states representations in pomdp," Internation conference on learning representations (poster) iclr 2014, 2014.

G. Dulac-Arnold, L. Denoyer, N. Thome, M. Cord, and P. Gallinari, "Sequentially generated instance-dependent image representations for classification," Internation conference on learning representations - iclr 2014, 2014.

L. Denoyer and P. Gallinari, "Deep sequential neural network," Arxiv preprint arxiv:1410.0510, 2014.

G. Dulac-Arnold, L. Denoyer, P. Preux, and P. Gallinari, "Sequential approaches for learning datum-wise sparse representations," Machine learning, vol. 89, iss. 1-2, pp. 87-122, 2012.
The ideal candidate will have:

  • A PhD on machine learning, data mining or other strongly related discipline.
  • A very solid background in a combination of computer science and mathematics. Special areas of interest include: statistical machine learning, reinforcement learning
  • Strong publication record in the area of machine learning
  • Solid expertise in at programming (mainly in C++).
  • Solid programming skills in scripting languages, such as python, etc.
  • Experience in deep neural networks and/or reinforcement learning if possible
  • Excellent command of English.
  • Team work capacity.

Candidates should send:

  • A two page CV.
  • A one page motivation letter explaining why their skills, knowledge and experience make them a particularly suitable candidate for the given position.
  • Their three most representative papers.
  • The contact details of two referees; do not send reference letters