I have a time series data that does not have an upper limit (data is somewhat monotonically increasing). Making the Test set values larger than the training set. (I am not shuffling because time sequence is important).
Following this, the test set should only be scaled with the training set's mean and standard deviation.
say I am scaling the training data to be between 0 and 1, won't this cause the test set to have values greater than 1? Isn't that a problem for the neural network if it has never trained on such values?