Describe fundamental principles of machine learning on Azure →mediumMultiple ChoiceObjective-mapped

AI-900 Practice Question: Describe fundamental principles of machine learning on Azure

What is 'model compression' and what techniques does it include?

⚠ Common exam trap

Test-takers frequently confuse model compression with general deployment or data optimization tasks, such as containerization (Option D) or data compression (Option A), because the word 'compression' is used broadly in Azure contexts.

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

✓

Reducing model size through pruning, quantisation, distillation, and factorisation for efficient deployment

Model compression is a set of techniques used to reduce the size of a trained machine learning model while preserving its accuracy as much as possible. This is critical for deploying models on resource-constrained devices like edge devices or mobile phones. The key techniques include pruning (removing unnecessary weights), quantization (reducing the precision of weights, e.g., from 32-bit floats to 8-bit integers), distillation (training a smaller 'student' model to mimic a larger 'teacher' model), and factorization (decomposing large weight matrices into smaller ones). Option B correctly lists these four core techniques.

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

✗
Compressing training data files to reduce storage costs
Why it's wrong here
Compressing training data files is a data-management practice that reduces storage cost and I/O, but it does not alter the trained model's architecture or parameter count. Model compression specifically targets the trained neural network—removing redundant weights or lowering numeric precision—so that inference becomes faster and less memory-intensive. The two operate on entirely different artifacts: raw datasets versus the learned model itself.
✓
Reducing model size through pruning, quantisation, distillation, and factorisation for efficient deployment
Why this is correct
Model compression reduces a trained model's computational footprint via pruning (removing low-importance weights), quantization (e.g., FP32→INT8), knowledge distillation (training a compact student to mimic a large teacher), and factorization (decomposing large weight matrices). These techniques enable edge deployment on resource-constrained devices and lower inference cost, usually with a small, controlled accuracy trade-off. This is precisely what the question describes.
✗
Summarising model documentation into a shorter model card format
Why it's wrong here
Summarising model documentation into a shorter model card addresses model governance and transparency, not the model's computational footprint. Model compression is an applied ML technique that shrinks the neural network through methods like pruning or quantization, whereas a model card is a non-executable artifact describing intended use and limitations. Condensing text cannot reduce inference time or memory usage.
✗
Packaging model code and dependencies into a container image for deployment
Why it's wrong here
Containerising model code and dependencies bundles the model into an executable package for consistent deployment, but it does not shrink the model itself. In fact, a container image often adds a runtime and libraries, increasing the overall storage footprint while preserving the original parameter count and precision. Model compression targets the model's internals; containerisation merely standardises the deployment environment.

Go deeper

AI-900 Practice Question: Describe fundamental principles of machine learning on Azure

Why each option matters

Option-by-option breakdown

Related to this question