We are looking for an excellent candidate for a Ph.D. position at the
Laboratory of Computer Science (LINA) of the University of Nantes,
France, starting from September 2013. The candidate will be part of the
TASC team at the LINA, one of the most leading research teams in the
field of Constraint Programming, producing among others the Choco
constraint solver.
*Title*: New parallel algorithms for Combinatorial Optimization
*Advisors*: Éric Monfroy and Florian Richoux
*Grant*: 3-year funding from the French Ministry of Research.
*Abstract*:
The evolution of computer architecture is leading us toward massively
multi-core computers for tomorrow, composed of thousands of computing
units. However nowadays, we do not know how to design algorithms able to
manage efficiently such a computing power. In particular, this is true
for combinatorial optimization algorithms, like algorithms solving
constraint-based problems. There exist several techniques for solving
constraint-based problems: Constraint Programming, Llinear Programming,
Boolean Satisfaction methods and Local Search methods to give an
non-exhaustive list. The latter is often among the most efficient
techniques to solve large size problems.
Nowadays and up to our knowledge, there exists only one algorithm
(Adaptive Search, a local search algorithm) showing very good
performances scaling to thousands of cores. However its parallel scheme
does not include inter-process cooperative communications. Moreover, the
rising of more and more complex algorithms leads to an number of
parameters which become intractable to manage by hand, and parallel
algorithms emphasize this trend.
In this context, this Ph.D. topic has two major objectives:
- To propose and implement local search methods scalable over thousands
of cores with cooperative and communicating processes. Beyond this
objective, there is another one: to define a portfolio approach,
having simultaneous processes of different nature (i.e., different
algorithms, or the same algorithm with different parameters) exploring
the search space.
- To develop and implement auto-tuning mechanisms for parallel local
search methods, i.e., mechanisms allowing to automatically manage
parameters (through machine learning techniques), and to include them
into local search methods developed during this thesis.
*Key words*: Constraint-based problems, parallel local search
algorithms, massively parallel architectures, auto-tuning, machine learning.
You can find the full Ph.D. topic at this address:
http://pagesperso.lina.univ-nantes.fr/~richoux-f/sujets/parallel_PhD-en.pdf
<http://pagesperso.lina.univ-nantes.fr/%7Erichoux-f/sujets/parallel_PhD-en.pdf>
**************************
*The salary *will be around *1420€ per month after taxe*, with benefits
such as health insurance and social coverage included. There are no
teaching duties, however the Ph.D.candidate will have the possibility to
teach (in French) at the University of Nantes. With teaching duties, the
salary will be around *1640€ per month after taxes*.
*Nantes* is the capital of the western French region Pays de la Loire.
It is the 6th largest city in France with 800.000 inhabitants, located
on the Loire River at 50km from the Atlantic Ocean. Nantes is one of the
most young and dynamic cities in France, and has been described in 2004
by the Times as "the most liveable city in Europe". The city is a hub in
Europe for IT innovations. This year, Nantes holds the title of European
Green Capital for its efforts to reduce air pollution and CO2 emissions,
for its high quality and well-managed public transport system and for
its biodiversity.
*The ideal candidate* should have a master degree or equivalent in
Computer Science or in Mathematics, and combine solid theoretical
background and strong programming skills (in particular C/C++).
Background knowledge and/or previous experience in the following areas,
though not mandatory, will be considered very favorably:
- For theoretical background: CSP, Meta-Heurisitics, Parameters
Auto-tuning, Machine Learning
- For software development: C++11, MPI or X10 programming.
If your are interested, please send your application in pdf as soon as
possible, with:
- a resume,
- a motivation letter,
- if possible, your Master grades,
- the name and contact of up to three references.
at the following addresses: eric.monfroy@univ-nantes.fr
<mailto:eric.monfroy@univ-nantes.fr> and florian.richoux@univ-nantes.fr
<mailto:florian.richoux@univ-nantes.fr>
Feel free to contact us for further information.
**********************************************************
*
* Contributions to be spread via DMANET are submitted to
*
* DMANET@zpr.uni-koeln.de
*
* Replies to a message carried on DMANET should NOT be
* addressed to DMANET but to the original sender. The
* original sender, however, is invited to prepare an
* update of the replies received and to communicate it
* via DMANET.
*
* DISCRETE MATHEMATICS AND ALGORITHMS NETWORK (DMANET)
* http://www.zaik.uni-koeln.de/AFS/publications/dmanet/
*
**********************************************************
Posts
