Proposta de um método de reamostragem em redes neurais com aplicação em transformação entre sistemas de coordenadas

Abstract

In developing artificial neural networks (ANN), the available dataset is split into three categories: training and testing. However, an important problem arises: How to trust the prediction provided by a particular ANN? Due to the randomness related to the network itself (architecture, initialization and learning procedure), there is usually no best choice. Considering this issue, we provide a framework which capture the randomness related to the network itself. The idea is basically to develop several networks for the same architecture and learning algorithm by repeating the random partitioning process of available data. Consequently, interval prediction is available instead of point prediction. The proposed method was applied to a coordinate transformation problem. The re-sampling method spatially characterizes the points of greater or lesser bias and accuracy. Although we restrict it to the neural network model, the resampling method proposed can also be extended to others modern statistics tools such as Kriging.