AI-900 Practice Question: Describe fundamental principles of machine learning on Azure
What is hyperparameter tuning in machine learning?
⚠ Common exam trap
Many exam-takers confuse hyperparameter tuning with model training itself (weight updates) or with data preparation steps (label correction, feature reduction), because all involve 'adjusting' something to improve accuracy, but only hyperparameter tuning searches over algorithm configuration settings that are set before training begins.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Searching for the best training configuration settings (learning rate, layers, etc.) to optimize model performance
Hyperparameter tuning is the process of systematically searching for the best combination of hyperparameters—such as learning rate, number of layers, batch size, or regularization strength—that control the training process itself, rather than being learned from data. In Azure Machine Learning, this is often automated using tools like HyperDrive, which runs multiple child runs with different hyperparameter configurations to find the set that maximizes model performance on a validation set.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Adjusting the training data labels to improve model accuracy
Why it's wrong here
Adjusting the training data labels refers to label cleaning or data correction, which is a data-quality activity that modifies the dataset itself to fix mislabeled examples. Hyperparameter tuning does not alter the training data; instead, it keeps the dataset fixed and varies the algorithm's configuration to find the best learning behavior. Label changes may improve accuracy indirectly, but they do not involve searching over training settings, so this option is incorrect.
- ✓
Searching for the best training configuration settings (learning rate, layers, etc.) to optimize model performance
Why this is correct
This option correctly defines hyperparameter tuning: it is a systematic search over training configuration settings—such as learning rate, number of layers, hidden unit sizes, regularization strength—to find the combination that optimizes model performance on validation data. Unlike model parameters, hyperparameters are set before training and are not learned from the data. Common tuning methods include grid search, random search, and Bayesian optimization, and the process aims to improve generalization by avoiding underfitting or overfitting.
- ✗
Reducing the number of features used by the model
Why it's wrong here
Reducing the number of features used by the model describes feature selection or dimensionality reduction, which is a data preprocessing step that changes the input space rather than the settings that control training. Hyperparameter tuning, in contrast, adjusts values like learning rate, batch size, and layer depth that govern the learning algorithm's behavior on that input space. While feature reduction can affect model performance, it is not a training configuration setting and therefore does not qualify as hyperparameter tuning.
- ✗
Updating model weights based on new production data
Why it's wrong here
Updating model weights based on new production data is model retraining or continual learning, where the learned parameters themselves are modified using fresh data. Hyperparameter tuning happens before or during the initial training process, and it selects the external configuration (e.g., learning rate, number of layers) that controls how weights are optimized. Retraining may be triggered by data drift and is a separate workflow from searching the hyperparameter space.
Go deeper
Related to this question
Learn chapter
Machine Learning Core Concepts
Key term
Azure Machine Learning
Azure Machine Learning is a cloud service for building, training, and deploying machine learning models at scale.
Key term
Model
In IT and AI, a model is a trained mathematical representation that learns patterns from data to make predictions or decisions.
About these practice questions
One of 985 original AI-900 practice questions on Courseiva, each with a full explanation and wrong-answer analysis — not exam dumps or protected exam content. Learn why practice questions differ from exam dumps →
JA
Written by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
This AI-900 practice question is part of Courseiva's free Microsoft certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the AI-900 exam.