Dataloader

[maizeng2008]

Hi, when you do the normalization for the event data, you also normalized the in_times, may I ask why?
For me, it is weird to normalize the actual time steps.
The mean_in and std_in is the mean of your time_step which is also a little bit weird.

The line of code I refered to is datasets.dataloader.py: line 153
self.in_times = [(t - mean_in) / std_in for t in self.in_times]

Also, what you are predicting is the inter-arrival time, although you did not use the inter-arrival time data to train the model but do we need to transform back the inter-arrival predicted data? I saw you directly use the prediction to calculate the MAPE metric.

Thank you