Measure whether the regression loss workbench works

Translate the claim into assertions or a tiny eval harness. The metric to protect is: MSE and MAE printed; removing the outlier changes MSE more than MAE. Always record the denominator (how many cases) beside any rate. A percentage without a denominator is marketing, not measurement.

def mae(y,p):
 assert len(y)==len(p)>0
 return sum(abs(a-b) for a,b in zip(y,p))/len(y)
assert mae([1,2],[1,2])==0
print(mae([2,4,6],[2,5,7]))

Expected evidence: MAE 0.667 and a passing exact-match test. A passing assertion proves only the behavior it names; broader usefulness still needs the chapter’s full limits.

Be explicit: beating the baseline (mean prediction residual before any model) on this fixture does not prove behavior under shape mismatch between y_true and y_pred, or silent NaN loss. Label observations separately from conclusions so the next page inherits honest evidence about the regression loss workbench.

Compute MSE and MAE printed; removing the outlier changes MSE more than MAE with the denominator written beside the rate every time. For this chapter, the evaluation set is intentionally tiny; that is allowed only if you say so in the evidence. Compare against mean prediction residual before any model before celebrating.

Add one negative case aimed at shape mismatch between y_true and y_pred, or silent NaN loss. A suite with only happy cases cannot protect the regression loss workbench when the characteristic failure appears in review.

If a check is expensive or flaky, shrink it until it is deterministic on y_true/y_pred arrays including one large outlier. Flaky green builds teach the team to ignore gates. Record what this page does not prove so security-ops and mastery-ship inherit honest limits.

At the evaluation stage for loss-functions, the job is narrower than finishing a product demo. You are creating one progressive evidence piece about y_true/y_pred arrays including one large outlier that later pages inherit without redefining success. Keep that fixture small enough to inspect by hand, keep outputs copy-pasteable as text, and refuse to narrate this baseline as if it were a production SLA: mean prediction residual before any model.

For this page specifically, success looks like metrics with explicit denominators and a negative case while still centering the user decision to compare MSE vs MAE on the same residual set to see which outlier hurts more. If you cannot point to a file, command, or assertion that proves that for the regression loss workbench, stay on this page instead of advancing.

Glossary: loss function · Glossary: gradient descent

Previous · Next

For loss-functions, write a transfer example that differs in one constraint from the chapter scenario. Keep the quality bar fixed. Explain which check still applies.