execution-plan/weeks/week-04-batchlens-first-release/materials.md · UPDATED 2026-10-01
Week 04 — Materials: what to read, what to watch
Status: PREPARED 2026-09-15. Links checked on 2026-09-15 unless marked (verify). Authority: v6 master plan, Week 4 "Learn it" block (units B4 microbial metabolism, M4 cross-validation), expanded into exact locators per the source contract.
Week 04 brief · Commitment · Evidence · Review · Gates · Previous
This is GATE 4 week. Reading is subordinate to the release. If the DOI is at risk on Friday, cut every optional item below and log the cut in review.md.
0101
Where this fits in the week
01
Where this fits in the week
| Slot | Budget | Material |
|---|---|---|
| Tuesday · B4 | 20 min | V3 Khan fermentation (10) → one paragraph in notes/bio/B4.md: "what is a fermentation, really". The rest of the block is build |
| Wednesday build block | 15 min max | R3 GitHub citation/Zenodo docs + the CITATION.cff format page |
| Thursday · M4 | 60 min + 30 | V1 (6) → R4 Kapoor & Narayanan, the leakage taxonomy only (20) → batch-wise CV with its test (~35); last 30 min founder paper |
| Saturday sandbox | optional | "Deeper", only after the release is out |
0202
Read
02
Read
| # | Source | Exact scope | Min | Why this week | Produce |
|---|---|---|---|---|---|
| R3 | GitHub Docs, "Referencing and citing content" and "About CITATION files"; format spec at citation-file-format.github.io | The Zenodo-integration steps and the minimal CITATION.cff fields |
15 | This is the literal procedure for Friday's DOI badge. Read it once, follow it exactly | CITATION.cff in the repo root and a rendering DOI badge in the README |
| R4 | Kapoor S. & Narayanan A., "Leakage and the reproducibility crisis in machine-learning-based science", Patterns 4(9):100804, 2023. DOI 10.1016/j.patter.2023.100804 · open PDF: hbiostat.org/papers/kap23lea.pdf | The taxonomy of eight leakage types; identify which one batch-wise CV prevents | 20 | The single best paper on the exact mistake that would invalidate Paper 2. Cite it in week 08 and week 18 | One paragraph in the README's honest "Where it fails" section naming the leakage type you guarded against |
0303
Watch (YouTube)
03
Watch (YouTube)
| # | Channel · title | ~Length | Link | When | Produce |
|---|---|---|---|---|---|
| V1 | StatQuest · "Machine Learning Fundamentals: Cross Validation" | ~6 min | youtube.com/watch?v=fSytzGwwBVw | Thursday, first | The sentence "why random k-fold is wrong for batch data" in notes/ml/M4.md |
| V3 | Khan Academy · "Fermentation and anaerobic respiration" (article + practice) | ~10 min | khanacademy.org/…/a/fermentation-and-anaerobic-respiration | Tuesday, first | One paragraph in notes/bio/B4.md: "what is a fermentation, really" |
0404
Deeper — Saturday sandbox only, and only after the release is out
04
Deeper — Saturday sandbox only, and only after the release is out
- OpenStax Microbiology, ch. 8 "Microbial Metabolism", sections 8.1-8.4 — start at 8.1 Energy, Matter, and Enzymes, chapter entry 8-introduction, recap 8-summary — 8.1 through 8.4; pay attention to aerobic vs anaerobic. Moved out of core on 2026-09-29.
- ISLP ch. 5 "Resampling Methods" — free PDF at statlearning.com — 5.1 cross-validation and 5.2 bootstrap; then the ch. 5 lab. Moved out of core on 2026-09-29.
- Sandve G. et al., "Ten Simple Rules for Reproducible Computational Research", PLOS Comp. Biol. 9(10):e1003285, 2013. DOI 10.1371/journal.pcbi.1003285 — All ten rules, they are one page each. Moved out of core on 2026-09-29.
- StatQuest · "Bootstrapping Main Ideas!!!" — Channel: youtube.com/@statquest — search the exact title. (verify): no stable video ID confirmed on 2026-09-15. Moved out of core on 2026-09-29.
- The Turing Way — book.the-turing-way.org, chapters on "Version control" and "Research compendia". The grown-up version of what you shipped this week.
- carpentries-lab/good-enough-practices — the hands-on lesson behind week 00's R3.
- Zenodo help — versioning and GitHub integration, if the DOI badge misbehaves.
- JOSS submission guidance — you submit in week 36, not now. Read it once so that today's release history is already the kind reviewers want.
0505
Link check log — 2026-09-15
05
Link check log — 2026-09-15
- Verified today: R1 (all three pages), R3 (GitHub docs, both pages), R4 (DOI + open PDF), R5 (DOI), V1, V3.
- Not verified: V2 — the StatQuest bootstrapping video did not resolve to a stable ID in today's search. Find it from the channel, record the URL you actually used in
evidence.md, or skip it; ISLP 5.2 covers the same ground. - Not re-checked today: R2 statlearning.com (verified 2026-09-12), citation-file-format.github.io, Turing Way, Zenodo help, JOSS docs.
- Gate 4 evidence must name the real DOI, the release tag and the command that regenerates the headline table. A prepared link is not a released artifact.
0606
NotebookLM kit sources for `LABLOOP-W04`
06
NotebookLM kit sources for `LABLOOP-W04`
(1) this file; (2) the Week 4 chapter from the v6 master; (3) R1 pages; (4) ISLP ch. 5 PDF; (5) V1 URL; (6) R4 PDF; (7) R5 article; (8) R3 GitHub docs pages; (9) your README draft with the headline number; (10) GATES.md.
Focus line: Focus: why a result is only finished when a stranger can regenerate it; batch-wise cross-validation as leakage prevention; aerobic versus anaerobic metabolism.
Custom question: Given my batch-wise CV implementation and the Kapoor & Narayanan taxonomy, which leakage type am I still exposed to in batchlens v0.1, and what is the cheapest test that would catch it?