Learn Before
Case Study

Choose the model that satisfies the time limit and gives the better score.

Case context: A team compares two text classifiers. Model A takes 145 ms per request and reaches 84% F1 score. Model B takes 110 ms per request and reaches 81% F1 score. The team treats running time as a satisficing metric with an acceptable limit of at most 150 ms.

Question: Using the satisficing-metric rule for running time, which model should the team choose, and why?

Sample answer: Both Model A and Model B meet the running-time limit of at most 150 ms, so running time no longer separates the two options. Because running time is only a satisficing metric, the team should choose based on the remaining metric, which is F1 score. Model A has the higher F1 score (84% vs. 81%), so Model A should be selected.

Key points:

  • Both models satisfy the 150 ms running-time limit.
  • A satisficing metric only needs to clear the threshold; it is not optimized further once the limit is met.
  • Since both models pass the time constraint, the choice should depend on the other metric.
  • Model A should be selected because it has the better F1 score.

Rubric: Full credit responses recognize that both models satisfy the running-time threshold, explain that running time is not the deciding factor after that point, and correctly choose the model with the higher F1 score. Partial credit for identifying that both models meet the time limit without making the final selection correctly.

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