Caracterização de Classes via Otimização em Redes Complexas

Lilian BertonLiang Zhao

Complex networks have emerged as an important way of data representation and abstraction able of capturing the topological structure presented in databases. This work proposes a method for building a network from a vector based dataset. It is based on the optimization of an energy function that considers purity and extension measures of the network. The constructed network was used to characterize mixing level among data classes in classification problem. Class characterization is an important issue, but it is not well studied. Therefore, we consider this work a contribution in this direction.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: