問題文
A single decision tree fitted on this data changes a great deal when a few rows are removed, and its score on unseen records swings from run to run. Which change reduces that instability the most?
選択肢
- The change that reduces it most is ensemble averaging in the design.
- The change that reduces it most is dimensionality scaling in the design.
- The change that reduces it most is hyperparameter freezing in the design.
- The change that reduces it most is gradient clipping in the design.