Plot: DistToNeighbor

[PhilippScheller]

1. Description Plot (min, max, mean) distance over the single iterations.

2. Open topics

   • [x] How to combine numeric and discrete parameters? Or should we have 2 separate plots for numeric and discrete params?

[juliambr]

• [x] Some wording: Always replace "features of the search space" or "features" by "search space components". Features ist natürlich nicht falsch, aber vielleicht ein bisschen verwirrend im Kontext von Hyperparameter-Optimierung.
• [x] gower --> Gower
• [x] "We use Gower distance to handle discrete search space components as well. "
• [x] In the formula, what is i, and j? Two different configurations? --> Definition is missing.
• [x] Describe also how min, mean, and max aggregation works (in more detail)
• [x] Small improvement of the plot: draw a vertical line at the position where the initial design ends (the number of points in the opt.path with dob = 0), usually 4 times the number of parameters. Then you can add some interpretation: You could see, whether the initial design should eventually be chosen bigger (e.g. if there is no significant drop after the initial design). Also, it seems that after the initial design has been created, no exploration happens any more - but this is just because of the scale. Maybe we can add an option to exclude the initial design points.