Configuration Values

Configuration values describe how an experiment should run.

Examples:

  • seed;
  • number of trials;
  • sample size;
  • train-test split size;
  • output path.

A dataclass can keep these values together.

Pass configuration into the experiment

Configuration controls the run

Runs locally with Python in your browser.

Ready to run.

The function does not hide its important choices. They are fields on SimulationConfig.

Configuration is not result

Keep configuration and output separate:

The config says what to do. The result says what happened.

Exercise: Configuration field

Which configuration field makes a randomized run repeatable?

Answer it first, then check.

Hint

Look for the field passed to np.random.default_rng(...).

Solution

The field is seed. It records the value used to initialize the random number generator.