StackedML
Practice
Labs
Questions
Models
Pricing
Sign in
Questions
/
Optimization
/
Hyperparameter Tuning
/
Hyperparameter tuning
← Previous
Next →
374.
Hyperparameter Tuning and Data Splits
easy
Why must hyperparameter tuning be performed on a validation set rather than the test set?
A
Using the test set for tuning leaks information about the test distribution, making the final evaluation optimistically biased
B
Using the test set for tuning violates the assumption that hyperparameters must be set before data is collected
C
Using the test set for tuning requires retraining the model from scratch after each configuration is evaluated
D
Using the test set for tuning inflates the variance of hyperparameter estimates since test sets are typically small
Sign in to verify your answer
← Back to Questions