User vector

[flyunicorn]

1. A bit confused of the lab here: 5.1 prediction for a new user and 5.2 prediction for an existing user. In 5.1 prediction for a new user, a user vector is provided and then user_vecs = gen_user_vecs(user_vec,len(item_vecs)). I guess this is to put the user feature provided through the neural network to generate the user vector with same length as movie vector. But in 5.2 prediction for an existing user, a different function is used: user_vecs, y_vecs = get_user_vecs(uid, user_train_unscaled, item_vecs, user_to_genre)
   What is the difference between the two?

2. Why we need to unscale y prediction?
   y_pu = scalerTarget.inverse_transform(y_p)

3. Comparing 5.1 with 5.2, two more lines are added in the end in 5.2 besides the different funciton used at the beginning. Why these two lines are needed in 5.2 not in 5.1? sorted_user = user_vecs[sorted_index]
   sorted_y = y_vecs[sorted_index]