ML platform
SageMaker Automatic Model Tuning
Runs hyperparameter optimization jobs.
Key points
- Runs multiple training jobs to search for better hyperparameter values.
- Optimizes a chosen objective metric such as validation accuracy or loss.
- Helps improve model performance without manual trial-and-error tuning.
- Works after an algorithm and training approach have been selected.
- Consumes training resources because each candidate configuration requires training work.
When to use it
- Choose Automatic Model Tuning when the model exists but hyperparameters need optimization.
- Use it to tune learning rate, tree depth, regularization, or other algorithm parameters.
- Use it when the question mentions objective metrics and many training jobs.
Exam tips
- Automatic Model Tuning tunes hyperparameters; Autopilot automates broader model building for tabular data.
- Tuning does not fix bad labels, leakage, or mismatched problem framing.
- Debugger inspects training behavior; tuning searches parameter combinations.
- More tuning can improve metrics but also increases cost and time.