ML Case Labs — Validation, Boosting, Segmentation, Pipelines
Back to M2 — Boosting + Calibrated Evaluation

Error-Analysis + Slice Writeup

3 error slices + fix proposals. Slice template inside.

71 minutesProject submission
🎯 Free Guest Mode: You are learning for free. Sign in to save your completion progress and quiz answers.

Assignment Instructions

Brief

Report 3 error slices on holdout predictions with a fix proposal per slice.

Dataset

Holdout predictions + labels; error-slice template provided.

Starter guardrails (88)

  • Template slots: slice definition -> size -> error rate vs base -> hypothesis -> fix.
  • Slices must cover >= 5% of holdout each (no anecdote slices).

Deterministic checks

Slice rates recompute from predictions; sizes meet the 5% floor.

100-pt rubric

  • Slice validity (definition + size): 25
  • Rate-vs-base analysis: 25
  • Hypotheses grounded in features: 20
  • Fix proposals (actionable, scoped): 20
  • Writeup clarity: 10
Checking your access…