How to get attention weights

[gugudexiatian]

Thank you very much for your answer. I now know how to run your code. But I have an another question: how do I get input attention weights as the authors did in the original paper ( A Dual-Stage Attention-Based Recurrent Neural Network for Time Series Prediction), I tried but failed. Looking forward to your help.

[Angelogeb]

Hi,

In order to obtain the input weights (or the temporal attention weights) you would have to save the node of the attention alpha (analogously beta for temporal attention) as an attribute of the model, i.e. refactor that line of code to self.alpha and then when doing session.run pass such node to be evaluated in addition to the rest.

Hope this is helpful!

Let us know if you have any additional question.

[gugudexiatian]

The last question, l have got it. Thanks! But l have another question. If I want to set the stride or horizon which is called in time series papers to 2,3,...7 rather than 1. I only changed the code in model.py by
for t in range(self.config.T):
if t<self.config.T:
if t<self.config.T-2 or -3 or -7 :
but the result is lagged. Maybe there is some wrong else or DA-rnn is not suitainable for prediction like this. I am really confusing. Because using the curent values to predict the target curent value is not significant in many cases.

[gugudexiatian]

Thanks for your amazing code and your share, but how to cite it?