Figure 4
From: A tabular data generation framework guided by downstream tasks optimization
Comparison of the downstream prediction Y and \(Y'\) across the California House datasets with the multiple targets. A perfect prediction (R2=1) is shown by the blue dashed line. The columns from 0 to 4 represent downstream target [’mean’, ’std’, ’max’, ’min’, ’MedHouseVal’] columns. Among them, ’MedHouseVal’ column is the target of downstream task prediction, while the others are constructed task targets.
