leanderweber

Change normalisation default, fix bug in normalise_by_negative, adapt citations, absolute imports

1

Changes of this PR: - [x] Make some Updates to the readthedocs - [x] Update and Unify Citations in all Metrics - [x] Update, unify, and sort all imports -...

Add a docstring or a warning for transparency

Due to patches being generated non-overlappingly, and depending on attributions, the number of patches per input can differ between samples, leading to different number of measurements between samples. While this...

### Description of the problem & solution Currently, the docstring of each metric mostly explains what the metric computes, as well as the corresponding reference. It may be helpful in...

attribution localisation metric should consider positive attributions only

1

### Description According to the paper (https://arxiv.org/pdf/1910.09840.pdf) attribution localisation should be computed only from the positive values in the attribution map. The implementation in quantus considers negative values also in...

### Description of the problem The accepted input shapes for segmentation masks (s_batch) seem very strict currently. E.g, for data of the shape (32, 3, 224, 224), only segmentation masks...

### Description of the problem Some metrics may output inf or nan values when inputting certain attributions. ### Description of a solution Such behavior should be caught and, at the...