Abstract: Despite of its slow learning time, back-propagation (BP) is one of the most widely used neural network training algorithms. In this paper, a nonlinear stretch method is presented that modifies the nonlinear activation function of BP algorithm to speed up the convergence. A invariant target recognition system based on BP neural network using this method and moment invariants is studied. Simulative recognitions on aircrafts and vehicles show that the speed of convergence is increased effectively.!8
展开▼
