Neural-Network Combination for Noisy Data Classification

Fabricio A. BreveMoacir P. Ponti Jr.Nelson D. A. Mascarenhas

Classifier combination experiments using neural network-based classifiers were carried out using noisy soil science multispectral images, which were obtained using a tomograph scanner. Using few units in the hidden layer images were classified by the Multilayer Perceptron (MLP) and the Radial Basis Function Network (RBF). Later we used classifier combining techniques as Decision Templates (DT) and Dempster-Shafer (DS), in order to improve the performance of the single classifiers and also to stabilize the performance of the Multilayer Perceptron. Classification results were evaluated through the Estimated Error (by the Hold-Out technique), and the Kappa Coefficient. The results showed that the RBF Network obtained a small improvement in performance with the combination. However, we observed a good improvement in stabilization of the Multilayer Perceptron, especially with the Decision Templates method.

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: