NarineK

@aarzchan, do you have a toy model demonstrating the use case ? IG computes the gradients w.r.t. the input tensors not the weight matrices and since `only_inputs =True` the gradients...

Thank you for the example, @aarzchan! Do you have Dropout or Batch Norm in the mode ? When you use Dropout, for instance, it is randomly choosing which neurons to...

GradientShap uses randomization. It selects baseline randomly and in addition to that it also randomly selects data points between input and baseline that's the only big difference compared to IG....

Thank you for raising the question, @jensqin! It looks like we could avoid softmax here. The documentation in PyTorch though calls softmax in the examples as well: https://pytorch.org/docs/stable/generated/torch.nn.CrossEntropyLoss.html?highlight=crossentropyloss#torch.nn.CrossEntropyLoss cc: @vivekmig

@ahmadajal, @vivekmig, is this still an open issue that needs to be addressed ?

@akashlp27, do you mind attaching your colab notebook here so that we can debug it ?

@basselmawzi, the weight vector is perpendicular to the hyperplane for the linear model by definition. https://www.cs.cornell.edu/courses/cs4780/2018fa/lectures/lecturenote03.html

Thank you for the tutorial, @JohannesK14! We will provide review comments soon! A quick question: Where is the `castle.jpeg` from ? We need to know the copyright permissions and the...

> Regarding your second comment: > > Of course, I can add the tutorial to the Resnet_TorchVision_Interpret.ipynb. But then I would suggest omitting VGG and just use Resnet and also...

@JohannesK14 would you have time to address the comments in this PR as discussed previously ?