Research of improved back-propagation neural network algorithm1

This paper pointes out the defects of practical application of BP(Back-propagation) algorithm. While this paper puts forward the concept of adaptive gradient factor on the base of the typical improved BP algorithms which other scholars presented, and puts forward a new BP improved algorithms with momentum term, adaptive gradient factor and adaptive learning step, obtaining the formula through deriving. The simulation experiments verify that the improved BP algorithm has some advantages in reaching high error precision, fast convergence speed and short recognition time.