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

Answer analysis

Option-by-option breakdown

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

• ✗

  The raw data files stored in Azure Blob Storage before any processing

  Why it's wrong here

  This option conflates storage objects with logical workspace resources. A raw Blob file or ADLS path is just an unstructured object in cloud storage; Azure ML datasets are abstractions registered in the workspace that add versioning, schema inference, a data profile, and access controls around that storage location. The same raw files can be registered multiple times as different versions, so the dataset is the reference, not the file itself. Blob storage alone lacks the lineage and sharing capabilities that make Azure ML datasets valuable.

• ✓

  Versioned, registered data references enabling reproducibility, sharing, and lineage tracking in Azure ML

  Why this is correct

  In Azure ML, a dataset is a versioned, registered reference to a data source, not a copy of the data itself. It encapsulates metadata such as schema, a data profile, and sampling strategy while pointing to files in a datastore like Azure Blob Storage or Azure Data Lake. This makes experiments reproducible because each run can pin a specific dataset version, and teams can share, re-use, and trace lineage across pipelines. The decoupling from model code is what enables these governance features.

• ✗

  Synthetic datasets automatically generated by Azure ML to supplement small training sets

  Why it's wrong here

  This option incorrectly equates Azure ML datasets with synthetic data generation, a technique that uses GANs, SMOTE, or other augmentation methods to fabricate training examples. None of these are automated by the Azure ML dataset registration feature; registered datasets are references to real, existing data sources, not algorithmically produced records. Azure ML does offer data labeling and data drift tools, but those operate on actual collected data rather than manufacturing new samples. Synthetic data generation would typically require separate libraries, not the Azure ML dataset API.

• ✗

  Pre-labelled benchmark datasets provided by Microsoft for testing Azure ML models

  Why it's wrong here

  This option mistakes Microsoft-provided sample or benchmark data for the Azure ML dataset concept. While Azure ML can access public datasets via Azure Open Datasets, the platform's own dataset feature is user-driven: you register and version your own data references in a workspace. Pre-labelled benchmarks like MNIST are commonly imported for demos but are not the definition of an Azure ML dataset. The core value being tested is reproducibility and lineage of customer data, not ready-made test corpora.

Go deeper

Related to this question