On hidden units and sample storing in multi-layered feedforward neural networks
Cui Rong · Journal of Yanbian University · 2001
In this paper, the mathematical descriptions of output and error vectors of least square approximation in three layer feedforward neural network model are introduced. Furthermore, the relation between real output, expected output and the error is discussed. The role of hidden units and the upper limit of the number of hidden units for storing samples are also analyzed.