Instrument the regression loss workbench

Instrumentation should explain behavior without collecting the raw material unnecessarily. For this artifact, record enough to calculate MSE, MAE, largest per-example contribution, and holdout error in the original target units. Include version or configuration identifiers so two runs can be compared. A log line is useful only if it answers a debugging question; dumping entire inputs creates noise and may create a privacy problem.

The artifact's user-facing goal is specific: compare mean squared error and mean absolute error on ordinary predictions and one costly outlier. Its accepted input is equal-length finite numeric target and prediction lists. Those statements are intentionally narrower than “build an AI system.” Narrow scope lets us inspect every input and expected result, and it prevents a toy result from being presented as a production claim. The sample emits a compact metric record that a test or dashboard could aggregate.

Save this as lesson.py and run python3 lesson.py. It uses only the language standard library, so the example is reproducible offline.

truth=[10,20,30]; pred=[11,18,36]
parts=[(p-y)**2 for y,p in zip(truth,pred)]
print({'mse':sum(parts)/len(parts),'largest_index':parts.index(max(parts))})

Expected output: MSE 13.667 and largest_index 2. Exact floating-point formatting may vary slightly, but the asserted behavior must not. Read the output as evidence about this stage, not merely proof that the interpreter started.

Print every residual and transformed contribution before averaging. A single large squared term should visibly dominate MSE; if it does not, verify operand order, exponent placement, and the denominator. Reject unequal lengths because Python's zip silently drops extras. Reject NaN and infinity before aggregation, and keep target units visible so a mathematically correct score is not interpreted in the wrong scale.

At the testing and observability stage, save the smallest failing fixture beside the expected result. Change one cause at a time and rerun the exact command printed above; that makes the repair reviewable and keeps this chapter's progressive artifact reproducible.

Compare MAE and MSE on the same holdout rows and show the largest contributor to each. Then translate the error into the product unit, such as minutes, dollars, or degrees. If underprediction costs more than overprediction, evaluate an asymmetric business cost separately; do not claim that minimizing a convenient textbook loss automatically minimizes operational harm.

For this testing and observability page, preserve the fixture and result as evidence for the next page. Label observations separately from conclusions: a passing assertion establishes the behavior it names, while broader usefulness requires the chapter's full evaluation set and stated operating limits.