Function approximation and time series prediction with neural networks

摘要: Neural networks are examined in the context of function approximation and related field time series prediction. A natural extension radial basis nets is introduced. It found that use an adaptable gradient normalized functions can significantly reduce amount data necessary to train net while maintaining speed advantage a linear weights. The local nature network permits simple learning algorithms with short memories earlier training data. In particular, it shown one-dimensional Newton method quite fast reasonably accurate