Machine Learning and Constraints

When:
30/06/2018 – 01/07/2018 all-day
2018-06-30T02:00:00+02:00
2018-07-01T02:00:00+02:00

Annonce en lien avec l’Action/le Réseau : aucun

Laboratoire/Entreprise : LIFO – Université d’Orléans
Durée : 3 ans
Contact : christel.vrain@univ-orleans.fr, thi-bich-hanh.dao@univ-orleans.fr
Date limite de publication : 2018-06-30

Contexte :
Machine learning and constraint programming are two important fields in artificial intelligence. Constraint programming offers a declarative and efficient paradigm to solve constraint satisfaction problems or constraint optimization problems. Machine learning and data mining problems are usually modeled as optimization problems or enumeration problems. Solving machine learning and data mining problems using constraint programming has recently interested both communities [4,6,1].

References
[1] M. Chabert and C. Solnon. Constraint Programming for Multi-Criteria Conceptual Clustering, CP 2017, pages 460-476, 2017.
[2] T.-B.-H. Dao, K-C. Duong and C. Vrain, Constrained Clustering by Constraint Programming, Artificial Intelligence Journal, vol 244, pages 70-94, 2017.
[3] T.-B.-H. Dao, C. Vrain, K.-C. Duong, I. Davidson. A Framework for Actionable Clustering Using Constraint Programming. ECAI 2016, pages 453-461, 2016.
[4] T. Guns, S. Nijssen and L. de Raedt. k-Pattern set mining under constraints. IEEE Transactions on Knowledge and Data Engineering, 2011.
[5] C.-T. Kuo, S. S. Ravi, T.-B.-H. Dao, C. Vrain, I. Davidson. A Framework for Minimal Clustering Modification via Constraint Programming. AAAI-17, 2017.
[6] J-P. Métivier, P. Boizumault, B. Crémilleux, M. Khiari and S. Loudni. Constrained Clustering Using SAT. In IDA 2012, LNCS 7619, pages 207-218, 2012

Sujet :

In this thesis, we are interested in developing declarative approaches using constraint programming for modeling and solving machine learning and data mining problems with structured data. More precisely, the data is not only defined by attributes or a distance measure, but also related by relations that define structures on data, as for instance graphs. One interest of a declarative approach is that the semantics can be integrated, as for instance labels on links, individual properties of instances or properties between instances etc. This work is based on our competences developed in LIFO on distance based constrained clustering using constraint programming [2,3,5].

Interested candidates are invited to send a CV, a motivation letter, the graduate level transcripts of marks as well as the name of reference persons.

The application must be done as soon as possible. The selected candidates will be invited to an interview.

Profil du candidat :
Machine Learning and Data Mining
Constraint Programming

Formation et compétences requises :
Master or engineering schools (including a reaserch training)

Adresse d’emploi :
LIFO – Université d’Orléans
Rue Léonard de Vinci
Orléans
France

Document attaché : Thesis-Orleans-2018.pdf