Essay

Use component error counts to set improvement priorities.

Question: A machine-learning team inspects 100 incorrect validation examples from a document-classification pipeline. They trace 84 errors to the document parser and 16 errors to the topic classifier. Explain how these counts should guide which part of the system receives the next round of engineering effort.

Sample answer: The team should use the error counts to identify the main source of failure. Because the document parser accounts for 84 of the 100 mistakes while the topic classifier accounts for only 16, the parser is the larger bottleneck. The team should therefore prioritize improving the parser, since fixing the component that causes most of the errors is likely to produce the biggest gain.

Key points:

  • Assign each validation mistake to the component that produced it.
  • Compare the counts across components.
  • Prioritize the component responsible for most of the errors.

Rubric: The response must state that error counts show which pipeline component is failing most often, identify the document parser as causing 84 errors and the topic classifier as causing 16, and conclude that improving the parser should be the priority.

0

1

Updated 2026-08-12

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Machine Learning Strategy

Machine Learning Yearning @ DeepLearning.AI