Sep 8, 2018 · 1 min read
Hey, I apologize for such a late reply. I must have missed the notifications. As for which x to use during after you have got the theta values for a new data point, you should use the original values. We normalize the data because we don’t want our theta values to get skewed. Once we do get the theta values there’s no point in normalizing new data that won’t be used for training but only for prediction. Of course the test and train data needs to be similar, otherwise it would be like predicting apples using theta values from training oranges.
Hope this helps. Over and out :-)
