Cumulative Review: Semesters 0–1
Budget 4–6 focused hours including correction. This is preparation for the independent exam, not another 50-question assessment. Draw on the launchpad's study/environment habits, Semester 0 code tracing, and Semester 1 mathematics.
| Retrieval task | Evidence to produce | Check or repair route |
|---|---|---|
| Negate a quantified API requirement | Domain and counterexample | Module 1 assessment key |
| Count labelled versus unlabelled assignments | Explicit equivalence relation | Module 2 counting contracts |
| Audit a hash-bucket claim | Dependence and deterministic bounds | E05 after your attempt |
| Interpret a confidence interval | Model, coverage and dependence caveat | Module 3 assessment |
| Separate residual from solution error | A two-coordinate numerical example | Module 4 numerical experiments |
| Read duplicate-sensitive search code | Trace, actual behavior and intended contract | Module 5 P15 |
| Connect graph counts to matrices | Triangle witness and trace calculation | Project fixture |
| Reproduce without the original session | Environment and command log | Semester 0 environment habits |
First work from memory; then check the linked module or key. For each missed task, write the exact missing assumption and solve a changed case. Reuse an existing valid artifact when it already answers the question; no copying quota.
Transfer before the exam
Explain how a proof obligation changes an implementation decision, how a data model changes a counting formula, and how a stopping rule changes an uncertainty claim. Record the two weakest areas and allocate repair time before the exam. Completion of this review alone does not pass the checkpoint.
Integer arithmetic evidence
Include the integer module gates and reuse the cycle audit. Recheck one signed Bezout certificate, one invalid cancellation and one complete solution list on a changed case. Explain the difference between logarithmic arithmetic-operation counts, bit cost and time proportional to the number of returned solutions.
Calculus evidence
Reuse the calculus audit and its delayed check. On changed cases, normalize a density, defend a constrained optimum, and separate a tail bound from quadrature error. Explain why numerical convergence alone does not prove a limit or continuity.
Probability inference retrieval
On changed cases, derive a transformed CDF, write a likelihood with units, update a stated prior using only new observations, and justify the permitted randomizations. Reuse the PS9–PS12 delayed checks. Explain why exact arithmetic cannot establish a valid sampling design.