Classification in Pattern Recognition. New Tools to Adapt a System to its Environment.
Synthèse des Approches de Classement en Reconnaissance des Formes. Nouveauxoutils pour l'Adaptation d'un Système a Son Environnement
OPEN ACCESS
This paper presents a new theoretic tool based on Information Theory, the main interest of which is to acutely evaluate the classification tools.
The particular nature of real-world objects recognition involves us to design systems based on multi-points of view approaches. The fusion stage has to adapt itself to the environment. We show that neural networks allow to learn the fusion function, optimized to the data and the structure of the composite system.
The performance of a composite recognition system is closed to the partition of the available information on each classification tools. A Genetic algorithm is designed to adapt the parameters space partition with the set of classification tools among the quality of the composite system, genetic algorithm.
Résumé
Cet article présente un nouvel outil théorique fondé sur la Théorie de l'Information afin de réaliser une évaluation d'un outil de classement plus fine que les mesures classiques.
Nous travaillons dans le cadre de la Reconnaissance d'objets naturels complexes et compliqués . La nature même du problème incite à travailler à l'aide d'une approche multi-points de vue décisionnels, fusionnés de façon adaptative. Nous montrons que les réseaux connexionnistes permettent l'apprentissage d'une fonction de fusion optimisée selon la nature du problème et la structure du Système de Reconnaissance.
Nous montrons aussi que la répartition de l'information sur chaque outil de classement contribue à une meilleure reconnaissance. Une approche de type génétique est alors conçue pour adapter la partition de l'espace des paramètres relativement à l'ensemble des outils disponibles.
Information theory, information fusion, multi-points of view, classifiers combination, neural network, pattern recognition, genetic algorithm.
Mots clés
Théorie de l'information, fusion d'informations, multi-points de vue, combinaison multi-classifieurs, réseau connexionniste, reconnaissance des formes, algorithme génétique.
[1 ] J. Albert, F. Ferri, J. Domingo, and M. Vincens. An approach to natural scene segmentation by means of genetic algorithms with fuzzy data. In Pattern Recognition and Image Analysis - 10(5),pages 97-113, september 1990.
[2] Ch. Amboise and G. Govaert. Spatial clustering and the em algorithm. Report, UTC Compiègne, february 1996.
[3] P. Andrey and P. Tarroux. Unsupervised image segmentation using a distributed genetic algorithm.Pattern Recognition, 27(5) :659-673, 1994.
[4] C.S. Beightler, D.T. Phillips, and D.J. Wilde. Foundations of optimization. Prentice-Hall, Englewood Cliffs, NJ, 2nd edition, 1979.
[5] I. Bloch. Fusion de données, ensemble flous et morphologie mathématique entraitement d'images.application à l'imagerie médicale cérébrale et cardiovasculaire multi-modalités. HabilitationàDiriger des Recherches 007, Ecole Nationale Supérieure des Telecommunications -GroupeImage, avril 1995.
[6] P . Bonelli and A. Parodi. An efficient classifier system and its experimental comparison with two representative learning methods on three medical domains. In4thInternational Conference on Genetic Algorithm,pages 288-295, may 4-8 1991.
[7] J. Branke. Evolutionary algorithms for neural network 'design and training . Technical Report 322, Institute AIFB, University of Karlsruhe, september 1995.
[8] N. Castignolles. Automatisation du comptage et de la reconnaissance des espèces dans les passes àpoissons par l'analyse de séquence d'images. Thèsededoctorat, Université Paul-Sabatier -Toulouse, janvier 1995.
[9] A. Colubi. Comparative studies of two diversity indices based on entropy measures : Gini-simpson's vs shannon's. In Information Processing and Management of Uncertainty in Knowledge Based Systems, pages 675-680, july 1996.
[10] I. Couso, P. Gil, and S.Montes.Measure of fuziness and information theory. In Information Processing and Management of Uncertainty in Knowledge Based Systems, pages 501-505, july 1996.
[11] H. de Garis. Modular neural evolution for darwin machines. InIJCNN, pages 194-197. Washingon, DC, 1990.
[12] K. De Jong and W.M. Spears. Learning concept classification rules using genetic algorithms. In11 th ICPR IAPR, pages 651-656. 12th International Joint Conference on Artificial Intelligence, 1992.
[13] A. DeLucaand S. Termini. A definition of anon-probabilistic entropy in the settingg of fuzzy sets theory. Information and control, 20:301-312, 1972.
[14] S. Deveughle.Etuded'unemithode decombinaisongraduelled'informations incertaines dans un cadre possibiliste. Thse, Université de Technologiede Compiegne, Compiegne, France, 1993.
[15] D. Dubois and H. Prade. Théorie des possibilités, applications à la représentation des connaissances eninformatique.Masson, Paris, 1988.
[16] D. Dubois and H. Prade. Combination of fuzzy information in the framework of possibility theory. InM.A.Abidi and R. C.Gonzales editors, Data fusion in Robotics and Machine Intelligence. Academic Press, pages 481-505, Boston, 1992.
[17] H. Dubois, D. et Prade. Théorie des possibilités : Applications lareprésentation des connaissances en informatique. InMasson, Paris, May 1985.
[18] H. Dubois, D. et Prade. La fusion d'informations imprécises. In Revue Traitementdu Signal Vol. 11, n°6, pages 447-458, 1994.
[19] A. Fioretto and A. Sgarro. A second step information measure and the uncertainty of bodies of evidence. InInformation Processing and Management of Uncertainty in Knowledge Based Systems, pages 687-691, july 1996.
[20] C.Fleurentand J.A. Ferland.Algorithmes génétiques hybridespourl'optimisation combinatoire. Technical report, Université deMontréal-Département d'Informatique et deRecherche Opérationnelle, 1994.
[21] R.G. Gallager. Information theory and reliable communication.John Wiley and sons, inc., 1968.
[22] G. Giraudon, P Garnesson, and P. Montesinos.Messie : un système multi spécialistes en vision. application à l'interprétation en imagerie aérienne. Traitementdu signal, 9(5) :403-419, 1992.
[23] D.E.Goldberg.Algorithmes génétiques. Addison-Wesley, 1994.
[24] MichelGrabisch. Characterization of fuzzy integrals viewed as aggregation operators. InIEEE Tron Fuzzy Systems, pages 1927-1932, 1994.
[25] M. Grabish. A survey of application of fuzzy measures and integrals. In5th IFSA Congress, pages 294-297, Seoul, July 1993.
[26] F.C. A. Groen, K. Ten, A.W.M. Smeulders, and I.T. Young. Human chromosome classification based on local band descriptors. Pattern Recognition Letters, 9(3) :211-222, 1989.
[27] Tahani H. and KellerJames M. Information fusion in computer vision using the fuzzy integral. InIEEE Tron Systems, man and Cybernetics, Vol.20, n°3, pages 733-741, May 1990.
[28] J. Heistermann. A mixed genetic approach to the optimization of neural controllers. In COMPEURO92: Computer Systems and Software Engineering, pages 459-464, may 4-8 1992.
[29] A. Hill and C.J. Taylor. Model-based image interpretation using genetic algorithm. InImage and Vision Computing,pages 295-301, june 1991.
[30] J.H. Holland. Adaptation in natural and artificial systems. University of Michigan Press, Ann Arbor, 1975.
[31] J.F. Jodouin.Les réseauxneuromimétiques.Modèleset applications.Hermès, 1993.
[32] L. N. Kanal. On pattern, categories and alternative realities. Pattern RecognitionLetters, 14(3) :241-255, 1993.
[33] A. Kaufmann. Introduction à la théorie des sous-ensembles flous -Applications à la linguistique, à lalogique et à la sémantique,volume 2. Masson, Paris, 1975.
[34] G.D. Kleiter and R. Jirousek. A maximum entropy approach for optimal statistical classification. In IEEE Workshop on Neural Network for Signal Processing, 1995.
[35] B. Kosko.Neural networks andfuzzy systems. A dynamical systems approach to machine intelligence.Prentice Hall International Editions, 1992.
[36] S. Kullback. Information theory and statistics.Peter smith edition, 1978.
[37] G.C. Langelaar and J.C.A. vanderLubbe. Adding inputs to existing neural networks. In Information Processing and Management of Uncertainty in Knowledge Based Systems, pages 1085-1090, july 1996.
[38] V Lefèvre, Y. Pollet, S. Philipp, and S. Brunessaux. Un système multi agents pour la fusion de donnéesen analysed'images.Traitementdu signal, 13(1) :99-112, 1996.
[39] K.S. Leung, Y Leung, and K.F. Yam. Rule learning in expert systems using genetic algorithm : 1, concepts. In 2nd International Conference on fuzzy logic and neural networks, pages 201-204. lizuka, Japan, july 1992.
[40] H. Li, Y . Gong, andJ.P. Haton. Apprentissagepar maximum d'information mutuelle pour des modèles neuronauxprobabilistes. Dizième Congrès Reconnaissance des Formeset IntelligenceArtificielle (Rennes), 2 :1043-1050, 15-19janvier 1996.
[41] P. Loonis. Contribution à la minimisation de l'a priorien Reconnaissance des Formes. Conceptiond'unprototype detrieuse automatiquedepoissons par VisionArtificielle en milieu industriel. Thèsede doctorat, Université de LaRochelle, janvier 1996.
[42] P. Loonis, M. Ménard, and J.P. Bonnefoy. Fusion d'information multisources : étude comparative entre une approche connexionniste dirigée et larègle orthogonalede dempster-shafer. Dizième Congrès Reconnaissance des Formes et Intelligence Artificielle (Rennes), 2 :606-614, 15-19 janvier 1996.
[43] P. Loonis, M. Ménard, and C. Demko. A new genetic algorithm for the multiclassifiers fusion optimization.Information Processing and Management of Uncertainty in Knowledge Based Systems, 2 :957-961, july 1996.
[44] R.J. Machado and A. F. da Rocha. Evolutive fuzzy neural networks. IEEE NN, pages 493-500, 1992.
[45] T. Matsuyama and V. Hwang. Sigma: a framework for image understanding . integration of bottom-up and top-down analyses. IEEE Transactions on Systems, Man and Cybernetics, 22(3) :908-915, 1992.
[46] D. Michie, D. J. Spiegelhalter, and C. C. Taylor. Machine Learning, Neural and Statistical Classification.Ellis Horwood, 1994.
[47] D. Miller, A. Rao, K. Rose, and A. Gersho. Learning bayesian networks under the conrol of mutual information. In Information Processing and ManagementofUncertainty in Knowledge Based Systems, pages 985-990, july 1996.
[48] T. Morimoto, J. De Baerdemaeker, and Y Hashimoto. Optimization of storage system of fruits usinf neural networks and genetic algorithms . In IEEE International Conference onFuzzy Systems,pages 289-294. Jarmo T. Alander, 1995.
[49] T. Murofushi and M. Sugeno. Fuzzy t-conorm integrals with respect to fuzzy measures : generalization of sugeno integral and choquet integral. InFuzzy Sets et Systems. Vol. 42, pages 57-71, 1991.
[50] M. Ramdani. Système dinduction formelle base de connaissances imprécises. PhD thesis, Université Paris 6, février 1994.
[51] M. Roux and J. Desachy. Information fusion for supervised classification in a satellite image. InProceedingsofthe Fourth IEEE International Conference onFuzzySystems. Vol. 3, pages 1119-1124, Yokohama, Japan, 1995.
[52] K. Sengupta and K.L. Boyer. Information theoretic clustering of large structural databases. IEEENN, pages 174-179, 1993.
[53] I. K. Sethi. Entropy nets:From decision trees to neural networks.Proceedings oftheIEEE, 78(10):1605-1613, 1990.
[54] G.A. Shafer. A mathematical theory of evidence. InNJ.-Princeton Univ.Press, 1976.
[55] G. Venturini. Apprentissage adaptatif et Apprentissage supervisé par Algorithme Génétique.Thèsededoctorat, Université Paris-sud,janvier1994.
[56] L. Wehenkel. On uncertainty measures used for decision tree induction. In Information Processing and Management ofUncertainty in Knowledge Based Systems, pages 413-418, july 1996.
[57] S.T. Wierzchon. On fuzzy measure and fuzzy integral. InFuzzyInformation and Decision Processes. Eds New York :North-Holland,pages 79-86, 1982.
[58] L. Xu, A. Krzyzak, and Suen C.Y Methods of combining multiple classifiers and their applications to handwriting recognition. IEEE Transactions on Systems, Man and Cybernetics, 22(3) :418-435, 1992.
[59] R.R. Yager. Connectives and quantifiers in fuzzy sets.FuzzySets and Systems, 40 :39-75,1991.
[60] L.A. Zadeh. Fuzzy sets as a basis for a theory of possibility. InFuzzy Sets ans Systems, 1, pages 3-28, 1978.