Comparison of training algorithms in neural network modeling of gamma spectrometric uncertainty

Abstract

A three-layer feed-forward artificial neural network with six different algorithms applied on different training sets was used to model uncertainties of activity levels of eight radionuclides (226Ra, 238U, 235U, 40K, 232Th, 134Cs, 137Cs and 7Be) in soil samples as a function of measurement time. The performance of applied neural network architecture is found to be very good, with correlation (R2) values between measured and predicted uncertainties ranging from 0.9291 for 7Be to 0.9915 for 137Cs.