论文部分内容阅读
BP is a commonly used neural network training method, which has some disadvantages, such as local minima,sensitivity of initial value of weights, total dependence on gradient information. This paper presents some methods to train a neural network, includi