Length of Pseudorandom Binary Sequence Required to Train Artificial Neural Network Without Overfitting

The artificial neural network (ANN) has been applied to the various fields due its capability process complicated nonlinear functions involving a large amount of data. A pseudorandom binary sequence (PRBS) is commonly used train ANN since PRBS easily generated by using linear feedback shift register and correlation function which peaked at zero delay but almost other delays. However, when length not sufficiently long (compared input size ANN), trained could suffer from overfitting where describes behavior training very well, does poorly on new data inputs. In this paper, we provide minimum required avoid for given ANN. For purpose, analyze estimate pattern through theoretical study. These analyses are confirmed numerical simulation. findings paper would be select