Quantus

Quantus copied to clipboard

### Description of the problem - Since the library is getting pretty big, it would be great to have proper logging configured. - This allows users to customize logs in...

aaarrti

### Description of the problem - First iteration of NLP support focuses on text-classification https://github.com/understandable-machine-intelligence-lab/Quantus/issues/103 - There is much more to NLP than text-classification, e.g., here is an overview of...

aaarrti

### Description of the problem - (https://github.com/understandable-machine-intelligence-lab/Quantus/pull/224) is designed with a goal to support models from Hugging Face Hub 🤗 because it is currently the most popular model sharing service....

aaarrti

### Description of the problem - Quantus is designed with a goal of being highly customizable. At this time, however, if user want to define his own, e.g., `similarity_func`, they...

aaarrti

### Description of the problem - Each metric has quite many parameters passed to `__init__`. Most of them are shared among all metrics, and play the role of configuration. ###...

aaarrti

### Description of the problem - Innvestigate is a popular library for explanation methods https://github.com/albermax/innvestigate ### Description of a solution - Build a wrapper around it - Related files: `explain_func.py`,...

annahedstroem

### Description of the problem - There are 6 ranking metrics presented in the paper: The Disagreement Problem in Explainable Machine Learning: A Practitioner’s Perspective https://arxiv.org/pdf/2202.01602.pdf ### Description of a...

annahedstroem

### 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...

leanderweber

### 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...

leanderweber

### Description of the problem - Create a category-wide normalisation option, to make it easier to do a comparison of metrics within the same category - Or add clarity if...

annahedstroem