What does hyperparameter tuning involve?
Show answer and explanation
Answer: B
Hyperparameters are configured before training and control how the learning process runs: learning rate, number of layers, tree depth, regularisation strength, batch size. Tuning searches combinations, often with grid, random or Bayesian methods, and compares validation performance to pick the best set. Parameters, by contrast, are learned from data during training. Tuning consumes compute, so a sensible search space and a solid validation strategy are essential.
Why the other options are wrong
- A. Feature selection decides the inputs, which is a data and feature engineering decision rather than a training setting.
- C. A learning rate schedule may change values during training, but tuning selects the configuration before training across candidate runs.
- D. Imputation repairs data quality and is unrelated to model configuration search.
Exam tip. Hyperparameters configure training; parameters are learned.