Question:
What characteristic makes an error category the 'most helpful' to include in your analysis?

Sample answer:
The most helpful error categories are the ones that you already have an idea for improving.

Key points:
- Having an idea for improving them

Rubric:
The answer must state that helpful categories are those for which the practitioner has an idea for improvement.

Google

The most helpful error categories are ones that you have an idea for improving. For example, the Instagram category is most helpful to add if you have an idea to undo Instagram filters and recover the original image. However, you do not have to restrict yourself only to error categories you know how to improve, because the goal is to build intuition about the most promising areas to focus on.

Most Helpful Error Categories Are Ones You Have Ideas to Improve

Machine Learning Yearning, a free ebook from Andrew Ng, teaches you how to structure Machine Learning projects.
https://www.deeplearning.ai/machine-learning-yearning/

Note: The content of the book is aligned with the Coursera Deeplearning.ai specialization.  https://www.deeplearning.ai/deep-learning-specialization/ 

Machine Learning Yearning (Deeplearning.ai)

Most Helpful Error Categories

Restricting Error Categories

Building Intuition on _____ Areas

Concepts in Error Category Selection

Process of Incorporating New Error Categories

Question:
Explain why it is beneficial to include error categories in your analysis even if you currently lack a clear idea of how to improve or fix them.

Sample answer:
Including error categories that you don't currently know how to improve is beneficial because the primary goal of the error analysis process is to build intuition. By tracking these errors, you can understand their frequency and impact on your model. This information helps you identify the most promising areas to focus your future efforts on, even if you don't have an immediate solution at hand.

Key points:
- Builds intuition
- Identifies the most promising areas to focus on
- Error categories do not require immediate solutions to be tracked

Rubric:
A strong answer should explain that tracking these errors helps build intuition and identifies promising areas for future focus, referencing that immediate solutions are not strictly required for a category to be useful.

Evaluating Unsolvable Errors

Case context:
You are developing an image recognition model and notice several misclassifications are due to Instagram filters. You aren't entirely sure how to fix this yet, but you think you might be able to find a way to reverse the filters.

Question:
Should you add an 'Instagram filters' category to your error analysis spreadsheet? Why or why not, based on the principles of categorizing errors?

Sample answer:
Yes, you should add the 'Instagram filters' category. The text explicitly states that an Instagram category is most helpful if you have an idea to undo the filters. Furthermore, even if you weren't completely sure how to fix it, you shouldn't restrict yourself only to errors you know how to improve, as tracking it will help build your intuition about whether it is a promising area to focus your time on.

Key points:
- Yes, the category should be added
- It is helpful because there is an idea to improve it (reversing filters)
- It helps build intuition about promising areas to focus on

Rubric:
A correct response will state that the category should be added and justify this by noting the user has an idea for improvement, and/or that tracking it builds intuition for future focus areas.

Learn Before

Related