PhD Position


Learning with non-stationary data - application to collaborative filtering and link prediction between name entities in knowledge bases like freebase


The continuous production of tremendous amount of data upsets the traditional view in science and information technology, particularly in machine learning (ML). These data evolve generally over time and, do not follow the fundamental hypothesis of stationarity upon which the learning theory is based. This is for example the case in collaborative filtering where the goal is to generate personalized recommendations for each user. Recommender systems filter out a potentially huge set of items, and extract a subset of N items that best matches user's needs with respect to other users preferences (observed) over existing items and who may have the same tastes than the latter. In this case, user preferences generally evolve over time ; as the perception of different items as well as their popularity are completely time dependent.
Learning in a non stationary environment, or learning concept drift, has found much attention in the ML community in recent years. Though learning algorithms in such environments have been formerly proposed, they were studied by making restrictive assumptions like, the partial availability of old data being generated with the past probability distribution, the impossibility of having new classes; and they have not been tested on non stationary applications.
The thesis aims at studying a new framework for this kind of learning and developing algorithms able to learn from large volumes of non-stationary data that come from real-life applications. We are particularly interested in learning problems such as collaborative filtering and link prediction in knowledge bases. Other related works, like zero-shot learning and transfer learning, are under investigation and the successful candidate will come to interact with other PhD and post-doc students working on these subjects.


For this position, we are looking for highly motivated people, with a passion to work in machine learning and the skills to develop algorithms for prediction in real-life applications. We are looking for an inquisitive mind with the curiosity to use a new and challenging technology that requires a rethinking visual processing to achieve a high payoff in terms of speed and efficiency. The applicant must have a Master of Science in Computer Science, Statistics, or related fields, possibly with background in machine learning, and/or optimization. The working language in the lab is English, a good written and oral communication skills are required.


The application should include a brief description of research interests and past experience, a CV, degrees and grades, a copy of Master thesis (or a draft thereof), motivation letter (short but pertinent to this call), relevant publications, and other relevant documents. Candidates are encouraged to provide letter(s) of recommendation and contact information to reference persons. Please send your application in one single pdf to The deadline for the application is May 15th, 2014, but we encourage the applicants to contact us as soon as possible. The final decision will be communicated in the beginning of June.
Duration: 3 years (a full time position) Starting date: September, 2014 Supervisors: Massih-Reza Amini (AMA, LIG) & Zaid Harchaoui (LEAR, INRIA)

Working Environment:

The PhD candidate will work at AMA team ( of the LIG lab and LEAR team ( of INRIA Rhone-Alpes at Grenoble. LIG ( and INRIA Rhone Alpes ( are leading institutions in Computer Science in France. Grenoble is the capital of the Alps in France, with excellent train connection to Geneva (2h), Paris (3h) and Turin (4h). AMA team is a dynamic group working in Machine Learning and connected scientific domains over 20 researchers (including PhD students) and that covers several aspects of machine learning from theory to applications, including statistical learning, data-mining, and cognitive science. LEAR team is a well-known computer science laboratory which main focus is learning based approaches to visual object recognition and scene interpretation, particularly for object category detection, image retrieval, video indexing and the analysis of humans and their movements.


Duration: 36 months
Salary after taxes: around 1597,11€,
Possibility of French courses
Help for housing
Participation for public transport
Scientific Resident card and help for husband/wife visa