神经网络总是提供相同的输出
I had to write a neural network, which could classify hand-written numbers from the MNIST data set. It seems to always predict the same class, however. I found this question, which describes a very similar case. However, the debugging methods, described in...