Learn Before
Case Study

Decide which metric to designate as the optimizing metric for a new classifier.

Case context: A team building a classifier has already decided that the classifier must run within an acceptable time limit, mirroring the running time criterion in the source's example. They now must decide how to treat their remaining performance metric, accuracy, when finalizing their evaluation approach.

Question: Based on the source's classifier example, what should the team do with accuracy once the running time criterion has been set, and why does this make accuracy the optimizing metric?

Sample answer: Since the running time criterion has already been set as a satisficing metric (a threshold to meet), the team should define accuracy as the optimizing metric and maximize it, subject to the classifier still meeting the running time criteria. This mirrors the source's guidance that after satisficing criteria are set, the final remaining metric becomes the one to maximize.

Key points:

  • Running time is already fixed as a satisficing threshold
  • Accuracy becomes the optimizing metric since it is the remaining metric
  • Accuracy should be maximized, not merely thresholded
  • The classifier must still meet the running time criteria while maximizing accuracy

Rubric: Full credit correctly applies the source's pattern: satisficing criterion set first (running time), then the remaining metric (accuracy) becomes the optimizing metric to maximize subject to the threshold. Partial credit for identifying accuracy as important without explaining the subject-to relationship.

0

1

Updated 2026-07-09

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