容错神经网络及容错BP算法研究

On a Fault-Tolerant Artificial Neural Network and Fault- Tolerant BP Algorithm

给出了容错神经网络的有关定义；建立了容错神经网络的数学模型，提出了容错神经网络的学习算法，即容错ＢＰ算法以及容错神经网络的结构设计方法；进行了仿真研究，并给出了实验结果及结论。

Some basic definitions of a fault-tolerant artifical neural network(FTANN)are givenand the learning promem of the fault-tolerant neural networkis reduced to a non-linear opti-mization one with the mathematical model developed.The learning algorithm,or the fault-tolerant BP algorithm,is proposed and the design method for the topological structure ofFTANN is discussed.Simulation on a PC computer has been carried Out.The fault-tolerantability of the neural network designed with all sorts of fault is discussed.Theoretical analy-sis and simulation results show that the fault-tolerant ability is good.