Machine Learning

Machine learning aims at acquiring knowledge by extracting patterns from raw data. The strong background of the ADRIA group on knowledge representation and reasoning naturally lead us to contribute to the extension of some the usual machine learning models to take into account uncertainty and imprecision in the data as well as in the models (Prade, Serrurier 2008 ; Serrurier, Raynal 2008 ; Serrurier Prade 2007ab ; Serrurier et al, 2007 ; Serrurier Prade 2006); to the study of machine learning as an argumentation problem (Amgoud, Serrurier 2008ab) and to learning to order things, both in a usual setting where the input data consists in examples of ordered pairs of alternatives, and in an often realistic setting where the input data consists in a list of alternatives that have been chosen / selected in the past (Chevaleyre et al, 2010 ; Fargier et al, 2019). We have also contributed to the application of Kolmogorov’s information theory for text classification (anti-spam filtering or classification of analogies): using information distance enabled us to use for that classical classification algorithms (Richard, Doncescu 2008 ; Belabbes Varzinczak, Richard 2008).

We have studied the theoretical aspects of analogical reasoning in order to use it as an induction tool. We are also interested in learning in the sense of Piaget’s theory. In this context, the thing learned is always constructed by transforming (“accommodating”) a pre-existing structure (“assimilating”). Thus, we do not learn new music insofar as it resembles other music already known. Within this framework, we have developed a program that teaches people to recognize musical rhythms in a natural and unsupervised way (Buisson, Quinton, 2009).

References

  • Leila Amgoud, Mathieu Serrurier. Agents that argue and explain classifications. Dans / In : International Journal of Autonomous Agents and Multi-Agent Systems, Springer, Vol. 16 N. 2, p. 187-209, mars 2008.
  • Leila Amgoud, Mathieu Serrurier. Arguing and explaining classifications. Dans / In : Argumentation in Multi-Agent Systems. Rahwan, Parsons, Reed (Eds.), Springer-Verlag, p. 164-177, Vol. 4946, LNAI, mars 2008.
  • Sihem Belabbes Varzinczak, Gilles Richard. Using SVM and Kolmogorov Complexity for Spam Filtering. Dans / In : International Conference on Artificial Intelligence (FLAIRS 2008), Miami (Florida), AAAI Press, p. 130-135, mai 2008.
  • Jean-Christophe Buisson, Jean-Charles Quinton. Internalized activities. Dans / In : New Ideas in Psychology, Elsevier, 2009.
  • Yann Chevaleyre, Frédéric Koriche, Jérôme Lang, Jérôme Mengin, Bruno Zanuttini. Learning Ordinal Preferences on Multiattribute Domains: the Case of CP-nets. Dans / In : Preference Learning. Hüllermeyer, Fürnkranz (Eds.), Springer, p. 273-296, 2010.
  • Hélène Fargier, Pierre-François Gimenez, Jérôme Mengin. Learning Lexicographic Preference Trees From Positive Examples. Dans / In : Proceedings of the 22nd AAAI Conference on Artificial Intelligence. AAAI Press, p. 2959-2966, 2018.
  • Henri Prade, Mathieu Serrurier. Bipolar version space learning. Dans / In : International Journal of Intelligent Systems, Wiley, Numéro spécial Bipolar Representations of Information and Preference (Part 2 : reasoning and learning), Vol. 23 N. 10, p. 1135-1152, 2008.
  • Mathieu Serrurier, Mathieu Raynal. Learning Fitts’ law with imprecise regression. Dans / In : Annual Meeting of the North American Fuzzy Information Processing Society (NAFIPS 2008), New York, IEEExplore digital library, p. 1-6, mai 2008.
  • Mathieu Serrurier, Henri Prade. Introducing possibilistic logic in ILP for dealing with exceptions. Artificial Intelligence>, Elsevier, Vol. 171, p. 939-950, 2007.
  • Mathieu Serrurier, Henri Prade. A general framework for imprecise regression. Dans / In : IEEE International Conference on Fuzzy Systems (FUZZ-IEEE 2007), London (UK), IEEE, p. 1597-1602, 2007.
  • Mathieu Serrurier, Didier Dubois, Henri Prade, Thomas Sudkamp. Learning fuzzy rules with their implication operators. Dans / In : Data and Knowledge Engineering, Elsevier, Vol. 60, p. 71-89, 2007.
  • Henri Prade, Mathieu Serrurier. Version Space Learning for Possibilistic Hypotheses. Dans / In : European Conference on Artificial Intelligence (ECAI 2006), Riva del Garda, Italy, G. Brewka, S. Coradeschi, A. Perini, P. Traverso (Eds.), IOS Press, p. 801-802, 2006.
  • Gilles Richard, Andrei Doncescu. Spam Filtering using Kolgomorov Complexity Analysis. Dans / In : International Journal of Web and Grid Services, Interscience Publishers, Vol. 4 N. 1, p. 136-148, 2008.