tiny bug ?

[christinazavou]

Hello,

thanks a lot for open sourcing your code for this interesting work!

I think there is a bug in this line: shouldn't it be output_freq = output_freq / torch.std(output_freq) ?

Thanks in advance,
Christina

[roseDwayane]

Thank you for pointing this out.

However, since the standard deviation of the normalized target signal is 1, lines 190 (output_freq = output_freq / torch.std(target_freq)) and 191 (target_freq = target_freq / torch.std(target_freq)) are redundant and do not have any function. These two lines have been removed.

Thank you again.

[christinazavou]

Ok I see thanks.

I also realized that in the simulated data, this line generates 19 identical signals for the 19 input/output channels. Was this on purpose or is it a mistake?

I also noticed using all the loss components:
trainValidateSegmentation(args=model_train_parameter([1, 1, 1, 1], './' + name + '_Simulate_5', "./" + name + "_simulate_data/"))
that loss1, loss2, loss3 lie within [0, 1] while loss4 is in the scale of hundreds.
Therefore I'm wondering if you used different data to generate the ensemble loss of figure 4?