Create concrete research tasks as issues: images

[elboyran]

Start with ideas from issue dianna-ai/dianna-exploration#187

[elboyran]

Created issues (to be addressed in roughly that order):

1. Image experiments: test shape images #470
2. Text experiements: RISE method parameters range dianna-exploration#172 (also relevant for text)
3. Experiements: LIME method parameters range dianna-exploration#173 (also relevant for text)
4. Experiments: KernelSHAP method parameters range dianna-exploration#174
5. Image Experiments (qualitative): LIME dianna-exploration#181
6. Image Experiments (qualitative): RISE dianna-exploration#180
7. Image Experiments (qualitative): KernelSHAP dianna-exploration#179
8. Image Experiments (quantitative): LIME dianna-exploration#178
9. Image Experiments (quantitative): RISE dianna-exploration#177
10. Image Experiments (quantitative): KernelSHAP dianna-exploration#176

[elboyran]

@geek-yang would you, please, have a quick look? It's not the best candidate issue for review, but a second pair of eyes is always good.

[geek-yang]

@geek-yang would you, please, have a quick look? It's not the best candidate issue for review, but a second pair of eyes is always good.

I quickly went through all the issues listed here. They all look good to me. There is only one thing I would like to add.

When addressing these research questions, it is good to document them in a way that could have added value to dianna. We can use notebooks to show the codes and results, and wrap-up our findings in markdown files. We can save them in the dianna-exploration repo, and later clean-up the repo and generate a sphinx doc to showcase what we've achieved. For instance, like Machine Learning 101 and Machine Learning Glossary.

This emphasizes the scientific value of using dianna. And it also makes life easier for @WillemSpek to wrap things up as a report 😉.

[elboyran]

Great suggestion, thanks @geek-yang!