Open rgtzths opened 3 years ago
I also have question about the padding is added at the end of the mains in WindowGRU code, it means window look at data after the target consumption time point. but in windowGRU paper
The technique we propose looks at a window of past aggregate data and infers the consumption at a single point. A window of length 𝑤 means that the neural network will receive the timeslice [𝑡-𝑤, 𝑡] to predict the device consumption at time 𝑡. An example can be seen in Figure 1.
The padding in the window-gru is added at the end of the mains instead of the beginning. Is this the supposed behavior?