Time traveling in time series models

I am a bit confused about when does time traveling happen in time series prediction. In the time series lecturer week, it is mentioned that "to not include data from the future in the training set. lets say I have collected data for 10 hrs, and at the time of serving I would use 2 hrs data to predict 1 hour into the future. Can I arrange the 10 hrs data I have collected into the following data and label sets, to train my model:

``````  X                    Y (labels)
``````

(t= 0, t = 1) (t = 2)
(t = 1, t = 2) (t= 3)
(t = 2, t= 3) (t = 4)
(t = 3, t = 4). ( t = 5) …

so basically a sliding window: is this considered time traveling or including data from the future in the training set ?

You feed in one data point from the future at a time.

The “sliding window” is created by the number of units in the LSTM when you created it.

Sorry so is this considered time-traveling or no?
In the above example
x_1 = (t= 0, t = 1) , y_1 = (t = 2)
x_2 = (t = 1, t = 2), y_2 = (t= 3)

if that is not clear, since I have (t=2) both as a label for one sample, and then part of the training data for another sample is this considered time traveling?
I am specifically confused about the “time traveling” cautionary note made in the lecture video in week 4.

Sorry, I do not understand your notation.

Can you give the video title and a time mark?