Fullscreen
Loading...
 
Print

Automatic parameter tuning job offers

2 PhD and 2 Post-doc positions are open in the area of automatic tuning of search and optimization methods in Orsay, France.

Context

Optimization and search methods for combinatorial or mixed discrete-continuous problems are reaching a mature state that allows users to tackle real-world problems in an efficient way. However, both exact methods (e.g. Constraint Programming) and heuristic and meta-heuristic approaches (e.g. Evolutionary Algorithms) have to face the critical issue of parameter tuning, that remains problem- and even instance-dependent, and requires past experience of the algorithm being used.
In this context, a project is being launched at the INRIA-Microsoft joint lab in Orsay, co-headed by Youssef Hamadi (Constraint Reasonning Group, Microsoft Research, Cambridge, UK) and Marc Schoenauer (Project-team TAO, INRIA Futurs, Orsay, France). Its goal is to set up automatic tuning methods for search algorithms in e-science, allowing scientists who have little knowledge of the search technique itself to nevertheless solve their optimization problem without the need for some "optimization engineer".
Relying on Machine Learning and statistical techniques, the project will address both off-line and on-line tuning issues, at the problem level as well as at the instance level. The target algorithms will be Constraint Programming, building on the expertise of the Constraint Reasoning Group at MSR, and meta-heuristics, with a particular emphasis on Evolutionary Algorithms, one research area of the TAO project-team at INRIA Futurs.
Two PhD and two post-doc positions are open (to start on October 1.). The location for all positions is the new INRIA-Microsoft Joint lab in Orsay, France . The PhDs? will be supervised by Youssef Hamadi (MSR), Marc Schoenauer (INRIA), and other TAO researchers. Needless to say, all the results of this project will be published and made available as Open Source.

PhD students

Profiles:

Both PhD students should have a Master in Computer Science or Applied Maths, or some equivalent diploma, with a strong background on Statistics and Machine Learning, and a solid programming experience. Some knowledge of optimization methods, either complete combinatorial methods or heuristic methods, is also mandatory. One position will be focused on Constraint Programming techniques, and the other one on Evolutionary Algorithms, but both should work together on the statistical learning techniques that are common to both problems when it comes to parameter tuning. Note that other related profiles will also be considered.
  • Competencies: Constraint Programming, Meta-heuristics, Machine Learning.
  • Technical skills: C, C++, C#, Matlab.
  • Additional competencies: French and English, Knowledge in Biology/Bioinformatics/e-Sciences, Development experience with MS Visual Studio.

Salary:

Standard French public PhD scholarship package.

Post-doc students

Profiles:

Post-doc candidates must have completed their PhD at the start of the project. The profile of the first Post-doc position is similar to that of the PhD students above. The other Post-doc should have a strong experience in bio-informatics, related to search or Machine Learning. She will work on the applications of the project, ensuring the practical relevance of the fundamental findings of the project.
  • Competencies: Constraint Programming, Meta-heuristics, Machine Learning.
  • Technical skills: C, C++, C#, Matlab.
  • Additional competencies: French and English, Knowledge in Biology/Bioinformatics/e-Sciences, Development experience with MS Visual Studio.

Salary:

Standard French public Post-doc package.

Contact:

Send full CV and at least two letters of reference to 'youssefh at microsoft dot com' and 'Marc dot Schoenauer at inria dot fr'.


Contributors to this page: evomarc .
Page last modified on Friday 25 of January, 2008 11:41:20 CET by evomarc.