%0 Conference Proceedings
%@isbn 85-244-0103-6
%T Um algoritmo geométrico para geração de funções discriminantes
%D 1996
%A Silva, José Ricardo Porto,
%A Tozzi, Clésio Luis,
%E Velho, Luiz,
%E Albuquerque, Arnaldo de,
%E Lotufo, Roberto A.,
%B Simpósio Brasileiro de Computação Gráfica e Processamento de Imagens, 9 (SIBGRAPI)
%C Caxambu
%8 29 out. - 1 nov. 1996
%I Sociedade Brasileira de Computação
%J Porto Alegre
%P 31-38
%S Anais
%1 SBC - Sociedade Brasileira de Computação; UFMG - Universidade Federal de Minas Gerais
%K cluster analysis, piecewise linear discriminant functions, pattern recognition.
%X Cluster analysis is a classification method which divides the pattern space into regions based on similary measure of the feature vectors. It is used for nonsupervised learning to further classification of inknown elements. There exist many clustering algorithms; in general they have limitations about clusters shape or demand the pre-definition of convergence parameters. This work presents a technique to quickly determine piecewise linear discriminant functions using geometric characteristics of subsets on the pattern space. Examples with synthetic and real data are included.
%9 Reconhecimento de Padrões
%@language pt
%3 a01.pdf