Loss functions

[PeterMcGor]

Hi all;

I have been reading the paper and I realised that the loss functions are not explicitly shown. Are you using the original loss functions described as in it architecture references?

Best!

[JoHof]

Hi Peter,

thanks for the comment. You are right we should have stated the loss used. Sorry about that. We used cross entropy

[PeterMcGor]

Thanks for the quick answer!

I just load your unet trained model to try TL and see that the last activation layer is LogSoftMax. Could be NLLLoss more appropriate?

Cheers!

[JoHof]

You can either remove the logsoftmax and use cross entropy or keep it and use nllloss. I guess it won't make much difference. I think I added the logsoftmax for this inference code only and used cross entropy for training

[PeterMcGor]

Sure! Just curious about stability and repeatability
Cheers!