Hello,

I am trying to develop a time series RNN. I use a time window of 8 vectors (t-7 to t) and my goal is to predict a sequence (t+1 to t+8) of regression values y for some chemical. The inputs x at every time step are vectors of size 6 of other chemicals, temp, pressures, etc. The inputs x and outputs y are not the same. I would have these x vectors of size 6 from t-7 to t. Since I am predicting a sequence of 8 values into the future, I will not know the values of the x vectors of size 6 after time âtâ. What input do I use for the RNN from time t+1 to t+8? Is there an option where the RNN simply does not have inputs after a certain point in time? I understand that I could simply give it input vectors of size 6 with random data in them or just zeros after time t. But is this a valid approach? I have been searching on line and have not been able to answer this question.

Thank you.

Hello,

Any help with this question?

Thanks