Learn Before
Adding a New Error Category During Review
Case context: You are evaluating a speech recognition system for customer-service calls. While sorting errors, you first tag problems as "low volume" and "overlapping speakers." Later, you notice many failures are actually caused by a new class: heavy background music.
Question: According to the iterative process used in error analysis, what should you do when you identify the "background music" category, and what about the examples you already reviewed?
Sample answer: Add "background music" as a new error category. Then revisit the earlier mistakes you already examined, because some of them may belong in this new group instead of the categories you assigned at first.
Key points:
- New categories can be added as they appear.
- Error analysis is iterative, not one-pass.
- Earlier examples may need to be reclassified after a new pattern is found.
Rubric: Full credit if the learner says to add the new category and to recheck the previously reviewed examples for reclassification.
0
1
Tags
Machine Learning
Deep Learning
Supervised Learning
Dive into Deep Learning @ D2L
Data Science
Machine Learning Strategy
Machine Learning Yearning @ DeepLearning.AI
Related
What best describes how error analysis should be carried out on a machine learning model?
True or False: You must decide every error label in advance before examining model mistakes.
Grouping mistakes can reveal patterns
Match each stage of an iterative mistake-review cycle to its description.
Arrange the stages of one error-analysis cycle in the correct order.
You find a new mistake pattern while reviewing model outputs for the first time. What is the best next step?
True or False: If error analysis uncovers a new failure type, earlier examples may need to be reviewed again under that new category.
After a second pass reveals new mistake patterns, inspect the same examples using the _____ categories.
Match each property of error review to the label that best describes it.
Arrange the steps that show why error analysis works best as a repeating process.
Why does error analysis often require repeated passes?
Adding a New Error Category During Review
Beginning error review without a checklist