The method of tolerance increasing to internal and external noises for neural network devices
S.N. Danilin, Mikhail V. Makarov, Sergey A. Shchanikov · 2014
We propose a method to reduce the impacts of noise distorting the input data on accuracy (quality) of the neural networks' operation. The essence of the method is to determine the necessary and sufficient digit capacity of input information and elements of neural networks. The offered method is realized in the form of the algorithm, whose work is shown by an example of some neural network devices.