How many validation errors should be reviewed by hand?
Question: Discuss the typical minimum number of validation mistakes that should be inspected manually and explain when reviewing many more mistakes is acceptable.
Sample answer: A common baseline is to inspect about 100 mistakes by hand. That is usually enough to reveal the main sources of error in the system. In some cases, analysts inspect many more, even around 500 errors, and that is fine as long as there is enough data to support the larger review.
Key points:
- Baseline of about 100 mistakes
- Usually enough to identify major error sources
- Larger reviews, such as 500 mistakes, can also be reasonable
- There must be enough data to support the extra review
Rubric: A strong answer should state the approximate baseline of 100 mistakes, explain that the purpose is to understand the main error sources, and note that reviewing more is acceptable when sufficient data is available.
0
1
Tags
Machine Learning
Deep Learning
Machine Learning Strategy
Supervised Learning
Dive into Deep Learning @ D2L
Data Science
Machine Learning Yearning @ DeepLearning.AI
Related
How Many Errors to Inspect Before Looking for Patterns?
Reviewing more misclassified examples is never useful once you already have a large dataset.
Baseline Number for _____ Misclassifications
Matching Error-Sample Counts to Meanings
Steps for Deciding on Manual Error Review Size
How many validation errors should be reviewed by hand?
How Many Model Errors Should Be Reviewed?
Main Value of Reviewing 100 Errors
When to Keep Reviewing Additional Errors
Checking About 100 Errors Gives a Useful Error Breakdown