Compare an analytical gradient with a finite-difference estimate before trusting a training loop.
Scenario
Loss is changing, but a sign or scaling error may still make the update mathematically wrong.
You will demonstrate
Project evidence
Each response is stored in this browser as you type. Include metrics, test output, or a decision rationale wherever the deliverable asks for it.
Record the scalar function, analytical gradient, numerical gradient, epsilon, relative error, tolerance, and result for one normal and one boundary case.
Required evidence: A two-row parity table plus the exact assertion used in the regression test.
0/80 minimum characters
Runnable Python lab
Implement a centered finite difference for a tiny scalar loss.
Project defense
Rate the evidence, not your effort: 0 missing, 1 weak, 2 adequate, 3 strong. All criteria must be reviewed, but a low honest score does not get hidden.
The check states epsilon and tolerance, compares independent calculations, and includes a boundary case.
Completion is controlled by stored evidence, deterministic tests, a decision defense, rubric review, and the artifact when required.
Optional cloud portfolio
An account is required. Submit only when the local completion gate passes. AI review is advisory and separate from deterministic completion.