采用贝叶斯正则化算法提高 BP 网络的推广能力。 在本例中，我们采用两种训练方法，即 L-M 优化算法（trainlm）和贝叶斯正则化算法（trainbr），用以训练 BP 网络，使其能够拟合某一附加有白噪声的正弦样本数据。-Bayesian regularization algorithm to improve the generalization ability of BP network.　 In this example, we use two training methods, namely LM optimization algorithm (trainlm) and Bayesian regularization algorithm (trainbr), to train BP network so that it can fit a sine additional white noise sample data.