Suggestion on the codes on page 419

[baifanhorst]

This is just a minor comment. Page 419 contains codes generating data for XOR classification. I do not understand why the numpy arrays are created first and then converted into tensors. I use ``X_train = torch.empty(N_train, 2).uniform_(-1, 1)'' to generate the features for the training set.

[rasbt]

I think it was just due to the conversion from the TensorFlow edition. I like your approach better too.