Deep Learning (DL) · Intermediate level

Intermediate level · Module 6: Intermediate Deep Learning Project · Lesson 2 of 3

Training and Comparison

Train candidate models with the project's training code and compare them against the targets.

In this lesson you will

  • Train and compare models reproducibly
  • Select a model against all targets

Key concepts

reproducible runscomparison tabletarget checkselection

The lesson

Use the configurable training code to train at least two architectures, logging every run.

Compare them in one table: validation accuracy, size, speed and training cost.

Select the model that meets all targets, not just the most accurate one.

Step by step

  1. Train the candidate models.
  2. Measure size and speed.
  3. Build the comparison table.
  4. Select and justify the model.

Practice activity

Complete the training and selection stage.

Expected result: A comparison table and a justified selection.

Check your understanding

  1. Why not simply choose the most accurate model?

    Show answer

    It may not meet size or speed targets for the phones it must run on.

  2. What should the comparison table include?

    Show answer

    Accuracy, size, speed and training cost.

Taught live by an instructor — online or in person. Free course lessons are for self-study and do not lead to a certificate — see registration options.