13.5 Chapter Summary

• regular and irregular grids, including space-filling designs

• build manually or using the family of grid_*() functions.

• tune_grid() can evaluate candidate sets of model parameters using resampling.

• autoplot() the tune object for the preferred performance metrics

• show_best() for a list of top models

• fast submodel optimization for some models / parameters on regular grids

• how to finalize a model, recipe, or workflow to update the parameter values for the final fit

• parallel processing backend capabilities

• consider racing methods to skip poor parameter combinations

Grid search is computationally expensive, but thoughtful choices in the experimental design can make them tractable.