Create notebooks

[Vayel]

• [x] Basic analysis, without confidence interval or categorical features
• [x] With categorical features
• [x] With confidence interval
• [x] Images
• [x] Comparison with Partial Dependence Plot (used in WIT: https://pair-code.github.io/what-if-tool/walkthrough.html)
• [ ] Comparison with Accumulated Local Effects

[MaxHalford]

I think that for the partial dependence plots, we can really highlight why our method works better. What what-if does, is that they shift a single feature and see what happens to the prediction. Naturally, because their model is linear they don't get the little kink when the age is above 60 years old. This in fact because of the fact that they are making hypothetical predictions with fake synthetic data. What we are doing, i.e. reweighting the data, ensures that we are not using fake synthetic data.