ABSTRACT

After the training process hundreds and even thousands of times, network error gets a stable network structure and connection weights and threshold of each neuron.

2 IMPROVING ALGEBRA ASSIGNMENT METHOD

Algebraic distribution law from American Electronic Equipment Reliability Advisory Group (AGREE) is proposed. It is a comprehensive consideration of the subsystem (unit), the complexity, importance and time difference, and a supposition that the subsystem failure rate is a constant, and independent

1 THE BASIC IDEA OF BP NEURAL NETWORK AND THE REALIZATION STEPS

BP (Back Propagation) network is a kind of error back propagation training algorithm for multilayer feedforward neural network; neural network model is one of the most widely used. Learning rules of BP network is to use the method of steepest descent, to constantly adjust the network weights and threshold by backpropagation, the network and the minimum sum of square error. As shown in Figure 1, the BP neural network topology includes input layer, hidden layer, and output layer.