Abstract: A method of estimating network reliability using neural network is proposed. Training it with the structure parameters, links reliabilities and exact reliabilities of serveral n node network models, the neural network can learn the relationship between structural parameters, links reliability and reliability of networks. Using the same training data the error estimation of the neural network model can be realized by means of GCV. To any network with n node, estimation of its reliability can be obtained by inputing its relative parameters into the neural network.