Case Study

Applying the data science analogy to model mistakes

Case context: Your image classification model is misclassifying a significant number of images. Instead of blindly trying a new algorithm, you decide to manually review a sample of the misclassified images to see if there are common characteristics among the errors.

Question: Based on the concept that error analysis is like data science for ML mistakes, what specific outcome are you trying to achieve by reviewing these misclassified images?

Sample answer: By acting as a data scientist analyzing the model's mistakes, the specific outcome you are trying to achieve is to derive actionable insights about what to do next, such as identifying which specific types of images are failing and focusing efforts on improving the model for those cases.

Key points:

  • The scenario describes error analysis.
  • It treats mistakes as data to be analyzed.
  • The outcome is to derive insights on what to do next.

Rubric: The student must identify that the outcome is to derive insights about the next steps to take for model improvement based on the analysis of mistakes.

0

1

Updated 2026-05-27

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