Learn Before
Decide When Incorrect Validation Labels Should Be Fixed
Case context: A team is comparing Model A and Model B using a held-out validation set. They discover that some validation examples are labeled incorrectly, and those mistakes are making it hard to trust the comparison and slowing down the review process.
Question: Based only on the decision rule in the lesson, what should the team do, and why?
Sample answer: The team should correct the mislabeled validation examples. The labeling errors are interfering with both the speed and reliability of the A-versus-B comparison, which is the situation in which fixing the labels is justified.
Key points:
- The validation set is being used to choose between two models.
- Wrong labels are making the comparison unreliable.
- The evaluation process has become slower or less decisive.
- Under these conditions, the labels should be fixed.
Rubric: The response should recommend fixing the labels and explicitly connect that choice to the fact that the model comparison is no longer fast or dependable.
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
Why mislabeled development examples matter more as models improve
When should mislabeled development-set examples be fixed?
When to Repair Validation Labels
Noisy labels can ____ the comparison itself.
Match each development-set concept to its role in deciding whether labels should be fixed.
Order the reasoning for deciding whether a noisy development set should be corrected.
When should mislabeled dev-set examples be fixed?
Decide When Incorrect Validation Labels Should Be Fixed
When is it worth fixing mislabeled validation examples?
When is cleaning dev-set labels worth the effort?
Fixing validation labels one by one is always required.