Milestone 3 of 8

Generate the repeatable observed data

Generate nine identified inputs with one local random generator, save exact variation and observations, and verify bounded same-seed replay.

The observed values should be reproducible without pretending that the added variation disappeared. Save the exact generated arrays and the rule that made them.

Goal

Generate the nine inputs, source-line values, bounded variation, and observed values with one local random generator. Save and replay the data from the configuration.

Inputs

Use the validated configuration, src/data.py, and one local generator created from the configured seed. Construct nine evenly spaced inputs from -2.0 through 2.0, calculate the source line, and draw variation from the half-open interval [-noise_bound, noise_bound).

The generated record must retain input, source value, variation, and observed value. The source slope and intercept are generation settings, not values that the project asks you to recover from the observations.

Deliverables

Implement a data-generation function in src/data.py that returns the four aligned arrays and the relevant generation evidence. Save output/generated_data.csv with one row per input and columns for input, source value, variation, and observed value.

Record the data identity, seed, input construction, and noise boundary in the report or the later manifest. The function must create its generator locally from the passed configuration; do not use a global random state.

Checks

Check that there are exactly nine inputs, that they are evenly spaced and include both endpoints, and that every observed value equals source value plus variation. Check finite values, the inclusive lower bound, and the exclusive upper bound of the supplied half-open interval.

Generate twice from the same configuration and compare all arrays. Generate with a changed seed and confirm that the recorded configuration explains the different variation without claiming that every value must differ. Check that the configuration and input arrays are not mutated.

Workspace

Keep data generation in src/data.py; use the accepted configuration from src/config.py. Write the CSV under output/ and keep its row order aligned with the input array. Do not place generated data beside the supplied plan.

Hints

HintSave the variation itself

Observed values alone cannot prove the noise bound. Keep the source value and variation so a later audit can reconstruct each row.

HintOne generator, one clear boundary

Create one generator from the seed and use it for the variation draw. The half-open interval matters when checking the bound.

Review

Open two rows in the CSV and calculate observed value from source value and variation. Explain what same-seed replay establishes and why a changed seed is only a controlled change in this supplied generator.

How to check your work

Checks compare the generated arrays and CSV columns with the supplied fixture. The supplied fixture preserves exact variation; it does not ask you to infer the hidden source line.