Microsoft Azure AI Fundamentals Machine Learning 5 — Questions and Answers
Question 1: Which Azure service provides pre-built ML models via REST APIs without requiring any model training?
- Azure Machine Learning Studio
- Azure Cognitive Services (Correct answer)
- Azure Databricks
- Azure Synapse Analytics
Correct answer: Azure Cognitive Services
Azure Cognitive Services offers pre-trained AI models for vision, speech, language, and decision-making, accessible via simple REST API calls without custom training.
Question 2: In the context of Azure ML, what is a 'model registry'?
- A list of approved Azure regions for deployment
- A centralized repository for storing, versioning, and managing trained ML models (Correct answer)
- A configuration file for compute clusters
- A catalog of available training datasets
Correct answer: A centralized repository for storing, versioning, and managing trained ML models
The Azure ML model registry is a centralized store that tracks all trained models with versioning, metadata, and lineage information for governance and deployment.
Question 3: Which machine learning technique combines predictions from multiple models to improve overall accuracy?
- Regularization
- Ensemble learning (Correct answer)
- Dimensionality reduction
- Hyperparameter tuning
Correct answer: Ensemble learning
Ensemble learning combines outputs from multiple models (e.g., bagging, boosting, stacking) to produce a stronger, more accurate prediction than any single model.
Question 4: What is the primary purpose of normalization or standardization in machine learning data preparation?
- Removing duplicate records from the dataset
- Scaling features to a similar range so no single feature dominates training (Correct answer)
- Converting text data to numeric format
- Splitting data into train and test sets
Correct answer: Scaling features to a similar range so no single feature dominates training
Normalization and standardization scale numeric features to comparable ranges, preventing features with large values from disproportionately influencing model training.
Question 5: In Azure AutoML for NLP tasks, which scenario type would you choose to assign a single category to each text document?
- Text NER
- Multi-label text classification
- Text summarization
- Multi-class text classification (Correct answer)
Correct answer: Multi-class text classification
Multi-class text classification assigns exactly one class label per document, whereas multi-label allows multiple labels; NER extracts entities rather than classifying documents.
Question 6: Which metric best evaluates a classification model when the classes in the dataset are highly imbalanced?
- Accuracy
- F1 Score (Correct answer)
- Mean Squared Error
- R-squared
Correct answer: F1 Score
F1 Score balances precision and recall, making it a better metric than accuracy when one class is much rarer, since accuracy can be misleadingly high by always predicting the majority class.
Question 7: What does 'hyperparameter tuning' mean in Azure Machine Learning?
- Adjusting the weights learned during model training
- Searching for the optimal configuration settings that control the training process (Correct answer)
- Cleaning outliers from the training dataset
- Selecting which features to include in the model
Correct answer: Searching for the optimal configuration settings that control the training process
Hyperparameter tuning (e.g., using Azure ML's HyperDrive) searches over settings like learning rate and tree depth that are set before training to find the combination that yields the best model.
Microsoft Azure AI Fundamentals Machine Learning 5
Which Azure service provides pre-built ML models via REST APIs without requiring any model training?