Plot Search Space

[PhilippScheller]

1. Description Plot search space for single features

2. Open topics

   • [x] partition plot area based on discrete/numeric features dynamically to ensure nice layout

[juliambr]

• [x] y-Achse: Value of search space component
• [x] The plot shows the values of the different search space components (y-axis) that have been evaluated by the optimizer over the number of iterations.
• [x] The color of the scatters corresponds to the value of the objective function y = f(x).
• [x] ... if the optimizer shows a positive ... it may be sufficient to limit the search space to larger (smaller) values in future runs.
• [x] Vielleicht hier kein xgboost Beispiel, sondern irgendwie allgemeingültiger formulieren

[juliambr]

• [x] another point: "prior" and "posterior" might be a bit misleading, because it reminds us of "true" Bayesian inference, which is not what we mean here. Maybe better "prior" --> "initial design sampling distribution"; and "posterior" --> "entire optimization run"