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.

• ✗

  A log of all Python exceptions and errors that occurred during model training

  Why it's wrong here

  Training exception logs are debugging tools that capture Python errors during the execution of model training code, such as syntax errors or invalid data types. Error analysis, however, is a model evaluation technique that finds data subgroups—like specific age brackets or geographic regions—where the model's prediction accuracy is notably lower than average. Reviewing exception logs tells you whether the training job ran correctly, not whether the model is underperforming on certain segments of the data.

• ✓

  Identifying data subgroups where the model makes disproportionately more errors than average

  Why this is correct

  Error analysis surfaces model blind spots by partitioning the dataset into cohorts based on input features, such as age, sex, or region, and comparing prediction accuracy across those cohorts. When a subset shows a disproportionately higher error rate than the overall test set, it indicates a systematic failure that aggregate metrics would obscure. This cohort-based inspection is exactly what Azure ML's error analysis dashboard is designed to reveal, supporting targeted model improvement and fairness review.

• ✗

  Counting the total number of incorrect predictions across the full test set

  Why it's wrong here

  Counting the total number of incorrect predictions on the full test set produces a single aggregate metric that ignores where those errors occur. Error analysis instead disaggregates performance by slicing the data into subgroups, so you can see whether errors are evenly distributed or concentrated in a particular demographic or feature range. Even a model with a low total error count could have a severe accuracy problem in a small but important cohort, and only a slice-by-slice comparison will surface it.

• ✗

  Reviewing error messages from failed Azure ML pipeline runs to diagnose infrastructure issues

  Why it's wrong here

  Pipeline failure diagnostics involve reviewing error messages from failed Azure ML runs to fix infrastructure issues like compute allocation, data access, or dependency problems. Error analysis is a post-training model quality tool that operates on the deployed or registered model's predictions, not on the running state of a machine learning pipeline. Infrastructure debugging answers the question 'Did the pipeline execute?' while error analysis answers 'Are the model's predictions fair and accurate across all data segments?'

Go deeper

Related to this question