Lesson 1 : Train_loss, valid_loss, error_rate

In lesson 1 we are doing multi-class classification so the the loss function(in this case train_loss and the valid_loss is CROSS-ENTROPY LOSS)
and the error rate is simply 1-accuracy.