Filtragem linear e abordagem geométrica baseada em grafos para o ajuste de parâmetros e redução de complexidade de redes neurais RBF
Co-author. This conference paper focus on learning without parameters. The algorithm uses information given by the Gabriel Graph structure and two filters, one based on the graph and another based on the LDA projection, to set the kernels of the RBF neural network. The scores are similiar to other parameterless methods but with, on average, less RBF centers.