Corrected "model.fit()" parameters

[mhmoodlan]

Just like #9 I've noticed that wrong parameters were passed to "model.fit()" in many notebooks: The 1st and 2nd parameters passed to "model.fit()" were the whole dataset (x, y, ...), whereas it should be only the training part (x_train, y_train).

List of affected notebooks:

1. t81_558_class5_class_reg.ipynb
2. t81_558_class6_backpropagation.ipynb
3. t81_558_class8_kaggle.ipynb
4. t81_558_class9_regularization.ipynb

[jeffheaton]

Thank you.