...
Université catholique de Louvain

Postdocs for project in machine learning - graph and networks - probabilities

Opslaan als favoriet E-mailalert

Over de werkgever

The Université catholique de Louvain (UCLouvain) is internationally recognized for research and teaching quality in many different fields of exper...

De pagina van de werkgever bekijken

We are looking for two excellent postdoctoral researchers to join us at the department of Mathematical Engineering, UCLouvain (Belgium) in the ongoing project

Learning from Pairwise Comparisons" (https://sites.uclouvain.be/learning_from_pairwise_comparisons/).

The positions can be filled immediately, and the project ends at the end of 2022 (subsequent opportunities to be discussed).

Project: The Learning from Pairwise Comparisons project focuses on learning problems in which a set of values have to be learned based on noisy information about pairs of these values (e.g. their ratio, product), a framework that applies to large classes of problems including for example

  • Recovering weights from comparison results in a Bradley-Terry-Luce model, with applications as diverse as modelling customer preferences, online advertisement, ranking quality of sport teams or evaluating patient reaction to medications
  • Determining experts' abilities without knowing the ground truth of their actions
  • General rank-1 matrix completion with perturbed data.

The project involves a collaboration with teams at Boston University and is funded by the Fund for Scientific Research (FRS-FNRS).

Objective and approach: We aim at developing generic ultra-rapid (near-linear) algorithms with minimax optimality guarantees for these learning problems. The project builds on preliminary works in which we re-formulated these problems as weighted least-square systems that are linear in the estimates but highly non-linear in the data (these works were published in ICML and AISTATS [1,2,3]). The data can be represented by a weighted graph in which two variables are connected when information on that pair is available, and this graph plays a major role both in the analysis of the algorithm and in the development of minimax lower bounds.

Possible objectives include

  • The development of online versions of these methods continuously incorporating new data,
  • The optimized selection of data to be queried, either a priori or in an online fashion, which is expected to be related to the graph properties,
  • The incorporation of prior information while keeping the algorithm efficiency.

Candidates: Ideal candidates have an expertise in at least one of the following fields: Machine Learning, Graphs and Networks, Probabilities, Online Learning, Optimization, Information Theory, and should be curious of new techniques. 
Candidates should hold a PhD which was obtained less than 6 years ago.

Timing: The position can be filled immediately. The project ends on Dec 31st 2022. (subsequent opportunities to be discussed)

Environment: UCLouvain is one of the oldest universities in Europe, with its main campus is located in Louvain-la-Neuve, close to Brussels, Belgium.

The Institute of Information and Communication Technologies, Electronics and Applied Mathematics (ICTEAM), hosts more than 40 professors and 200 researchers, and carries out both basic and applied research in key fields of information and communication technologies, electronics, computer science and applied mathematics. Its members teach in the Ecole Polytechnique de Louvain.

You will be joining a team of currently 8 researchers, including one postdoc working on the project.

How to apply: Send an email with title “[Postdoc Pairwise]” to Julien Hendrickx (julien.hendrickx@uclouvain.be), containing a CV, list of publications, and the name of at least two referees. There is no deadline, the positions remain open until filled.

If you have any question, please contact us at julien.hendrickx@uclouvain.be.

Links

PI: Julien Hendrickx: https://perso.uclouvain.be/julien.hendrickx/

Project: https://sites.uclouvain.be/learning_from_pairwise_comparisons/

ICTEAM/INMA: https://uclouvain.be/en/research-institutes/icteam/inma

Ecole Polytechnique de Louvain (EPL) : https://uclouvain.be/en/faculties/epl

UCLouvain: https://uclouvain.be/fr/index.html

F.R.S-FNRS (fund for scientific research): https://www.frs-fnrs.be

Louvain-la-Neuve : https://en.tourisme-olln.be/

References

1. Julien Hendrickx and Alex Olshevsky: “Minimax rank-1 matrix factorization”, 
In Proceedings of AISTATS, the International Conference on Artificial Intelligence and Statistics (AISTATS2020), 2020.

2. Julien Hendrickx, Alex Olshevsky, and Venkatesh Saligrama: “Graph resistance and learning from pairwise comparisons”, In Thirty-sixth International Conference on Machine Learning (ICML2019), 2019.

3. Julien Hendrickx, Alex Olshevsky, and Venkatesh Saligrama: “Minimax rate for learning from pairwise comparisons in the BTL model”, In Thirty-seventh International Conference on Machine Learning (ICML2020), 2020.

ADDITIONAL INFORMATION

Benefits

Stipend of 2550-2727€/month free of tax. Benefits include free public transportation for the commute and coverage by the Belgian Healthcare System. Funding is available for conferences and travels in the context of collaborations.

Eligibility criteria

Candidates must hold a PhD obtained less than 6 years prior to the date of hiring.

Selection process

To apply, please send an email with title “[Postdoc Pairwise]” to Julien Hendrickx (julien.hendrickx@uclouvain.be), containing a CV, list of publications, and the name of at least two referees.

The positions remain open until filled.

REQUIREMENTS

Offer Requirements

Skills/Qualifications

Ideal candidates have an expertise in at least one of the following fields: Machine Learning, Graphs and Networks, Probabilities, Online Learning, Optimization, Information Theory, and should be curious of new techniques.

Nu solliciteren

Persoonlijke gegevens*
Je cv en bijlagen uploaden

*Door te solliciteren op een functie op Academic Positions ga je akkoord met onze algemene voorwaarden en ons privacybeleid.

Informatie over de vacature

Functienaam
Postdocs for project in machine learning - graph and networks - probabilities
Locatie
Place de l'Université 1 Louvain-la-Neuve, België
Gepubliceerd
2021-11-12
Uiterste sollicitatiedatum
2021-12-11
Soort functie
Opslaan als favoriet E-mailalert

Meer vacatures bij deze werkgever

Over de werkgever

The Université catholique de Louvain (UCLouvain) is internationally recognized for research and teaching quality in many different fields of exper...

De pagina van de werkgever bekijken

Vergelijkbare vacatures

...
Become an IB Examiner for Swedish A: Literature International Baccalaureate® (IB) 1 week geleden
...
Solar Energy Assistant/Associate Professor University of Nizwa 10 maanden geleden
...
Assistant Professor in Marketing Wenzhou-Kean University 3 maanden geleden
...
Become an IB Examiner for Croatian A: Literature International Baccalaureate® (IB) 1 week geleden
...
Become an IB Examiner for Danish A: Literature International Baccalaureate® (IB) 1 week geleden
Meer vacatures

Relevante stories

...
The Power of Ultrafast Lasers Advanced Research Center for Nanolithography ARCNL Leestijd: 4 min
...
Going Nuclear to Save Time University of Jyväskylä Leestijd: 5 min
...
Towards a Green and Ethical Future University of Jyväskylä Leestijd: 4 min
...
Herbal Warfare: A Natural Weapon to Fight Parasites Free University of Bozen - Bolzano Leestijd: 4 min
Meer stories