In the last few years we have been developing a series of models for classification and regression which are based on the geometrical structure of the data set. Particularly, we have concentrated our effors in the construction of a Gabriel Graph, since they have interesting properties that make its application to Machine Learning problems attractive. Large margin models for classification and for regression can be obtained without parameters provided by the user, so the models that we obtain are parameterless and appropriate for online and incremental learning. In addition integrated circuit implementation is feasible. Our main publications in the area are listed below.
Torres, Luiz C. B.; Castro, Cristiano L. ; Braga, Antônio P.. A Computational Geometry Approach for Pareto-Optimal Selection of Neural Networks. Lecture Notes in Computer Science. 1ed.Berlin: Springer Berlin Heidelberg, 2012, v. 7553, p. 100-107.
TORRES, L. C. B.; COELHO, Frederico ; CASTRO, C. L. ; BRAGA, A. P.. A Graph of Gabriel Approach for Large Margin Classifiers. In: LA-CCI - The Latin American Congress on Computational Intelligence Co-located with ARGENCON, 2014, San Carlos de Bariloche. Proceedings LA-CCI 2014, 2014. v. 1. p. 25-29.
Torres, Luiz C. B.; Lemos, André P. ; Castro, Cristiano L. ; Braga, Antônio P.. A Geometrical Approach for Parameter Selection of Radial Basis Functions Networks. Lecture Notes in Computer Science. 1ed.: Springer International Publishing, 2014, v. 8681, p. 531-538.
TORRES, L. C. B.; Castro, C.L. ; BRAGA, A. P.. Gabriel Graph for Dataset Structure and Large Margin Classification: A Bayesian Approach. In: Proceedings of the European Symposium on Neural Networks, 2015, Bruges. ESANN 2015, 2015. p. 237-242.
TORRES, L.C.B.; Castro, C.L.; COELHO, F. ; SILL TORRES, F. ; BRAGA, A.P.. Distance-based large margin classifier suitable for integrated circuit implementation. Electronics Letters, v. 51, p. 1967-1969, 2015.