In PhilippScheller/VisBayesOpt: Visualizing model based optimization runs from package mlrMBO

MboPlotOptPath

Description
The plot shows the surrogate model for the chosen iteration.

Interpretation
* 1D search space: Shows the graph of the surrogate model and the acquisition function (infill criterion).
* Higher dimensional search space: Shows the Partial Dependence Plot PDP of the surrogate model with regards to the chosen search space component. At iteration $i$, the marginal effect of the chosen search space component on the predicted outcome $\hat y^{(i)}$ of the surrogate model at iteration $i$ is computed and displayed.

How to use
* Identify the influence of a search space component on the prediction of $\hat y^{(i)}$: A partial dependence plot can show whether the relationship between the target and a search space component is linear, monotonic or more complex.

Modifications * $highlight_iter$: We can set $highlight_iter$ in the sidebar panel if we are interested in a specific iteration to be highlighted.
* $feature$: We can select a search space component ($feature$) to inspect the surrogate (or PDP for higher dimensions) for that component.

PhilippScheller/VisBayesOpt documentation built on Sept. 14, 2020, 12:47 p.m.