Plots

[hageldave]

Plots to be implemented

Plots for collections of distributions

These plots eat a list of distributions and depict them jointly.

1D plots

any of the 1D plots should also be applicable to nD distributions where each dimension is treated independently (e.g. only looking at the diagonal of the covariance matrix for a multivariate normal distribution), and will then produce n such plots for each dimension.
For example a box plot showing 3 distributions where each distribution has 6 dimensions (HP, Att, ...):

• Box Plots
• Violin Plots
• Dot Plots

2D plots

• scatter plot (using samples drawn from the distributions)
• contour plot (showing isolines of the probability density functions)
• contour bands
• stippling of PDFs e.g. multi-class inverted stippling (Schulz et al.)

Plots for dimensionality reduction evaluation

• Shepard diagram
  • highdim vs lodim samples
  • highdim metrics on distribution vs lodim metrics (e.g. pairwise distance of means, or trace(covA+covB))