rachtibat

I love you all! Thanks, I had the same issue

Yes, I can confirm this. Good change!

Hi, @bugsuse I see that you use the SequentialMergeBatchNorm canonizer. As far as I understand, this canonizer is only used for BatchNorm Layers. Does your UNet implementation contain torch.add layers...

@bugsuse Hi, maybe you can try out the new pull request #45 with `git fetch origin pull/45/head:YOUR_BRANCH_NAME` This produces much better heatmaps with LRP

Hey, Thank you Rodrigo for putting so much effort into solving the instability and thanks to Christopher for taking the time to answer all the issues in depth. Although I...

Hi Chris, Thanks that's interesting. It confirms the correlation between heatmap quality and hidden concept attribution.

Regarding the second feature "detecting unexpectedly changed relevance values", this could be implemented by introducing a new rule/hook that 1. returns fixed relevance values, e.g. torch.ones_like(input.shape) * 42. 2. and...

Hey, awesome many thanks for the detailed explanation. I am very excited about the future development and will have a look at the PR to see if I can modify...

Hey, thank you for your prompt and thoughtful response as always. I like the idea to add a `FuncMod`. I ask Sebastian, and he told me that another use-case would...

Hey, sorry for the late reply.