Saving a Summary

A summary records the run.

For this small experiment, save:

  • seed;
  • number of points;
  • true slope;
  • fitted slope;
  • loss;
  • one written observation.

Write a summary file

Save experiment summary

Runs locally with Python in your browser.

Ready to run.

The file is plain text so the result is easy to inspect.

Summary is not a diary

A useful summary is short. It records what someone needs to understand or rerun the experiment.

If every number is saved but the seed is missing, the summary is incomplete. If the plot is saved but no observation is written, the inspection is unfinished.

Exercise: Summary field

Which field is essential for rerunning the same noisy data?

Answer it first, then check.

Hint

Record the value passed to np.random.default_rng(...).

Solution

The essential field is the seed. Recreating the same run also requires the same code, data-generation settings, and relevant environment.