The course at a glance¶
A 13-session semester (13 × 180 min), worth 6 ECTS: one taught precourse session covering both precourse decks, then twelve chapter lectures over ten ISLP chapters.
13 sessions × 180 min 6 ECTS 1 taught precourse session 3 decks split across two lectures 3 + 5 practice exams
Semester plan¶
The precourse opens the semester as a taught session; the twelve chapter lectures keep their numbers 1–12, which is what the exam calendar and the runsheets refer to.
Session |
Chapter(s) |
Content |
Exercises |
|---|---|---|---|
Precourse |
Ch 0 + Ch 0b |
Taught in one session, drawing on both precourse decks: descriptive statistics; probability and Bayes; distributions; standard errors and confidence intervals; hypothesis testing; simple linear regression; the Python toolkit — and reading mathematical notation; logs and exponentials; odds, the logit and the sigmoid; likelihood; counting and computational cost (with a lab notebook each) |
0.1–0.10, 0b.1–0b.6 |
1 |
Ch 1 + Ch 2 (part 1) |
Introduction; what is statistical learning; prediction vs. inference; parametric vs. non-parametric |
1.1–1.3, 2.1–2.2 |
2 |
Ch 2 (part 2) |
Assessing accuracy; bias–variance trade-off; classification & KNN; lab |
2.3–2.8 |
3 |
Ch 3 (part 1) |
Simple & multiple linear regression; estimation; inference (SE, t, F) |
3.1–3.6 |
4 |
Ch 3 (part 2) |
Qualitative predictors; interactions; diagnostics; KNN vs. OLS; lab |
3.7–3.12 |
5 |
Ch 4 (part 1) |
Logistic regression; odds; multiple logistic regression; confounding |
4.1–4.4 |
6 |
Ch 4 (part 2) |
The confusion matrix; ROC/AUC; lab — the generative models (LDA, QDA, naive Bayes) are appendix material |
4.8–4.10 |
7 |
Ch 5 |
Validation set, LOOCV, k-fold CV, the bootstrap |
5.1–5.6 |
8 |
Ch 6 |
Subset selection; Cₚ/AIC/BIC; ridge; lasso; PCR/PLS |
6.1–6.7 |
9 |
Ch 7 |
Polynomials, step functions, splines, smoothing splines, GAMs |
7.1–7.6 |
10 |
Ch 8 |
Decision trees; bagging & OOB; random forests; boosting |
8.1–8.7 |
11 |
Ch 10 |
Neural nets; forward pass; backprop/GD; CNNs; regularization (PyTorch) |
10.1–10.6 |
12 |
Ch 13 |
Multiple testing; FWER; Bonferroni; Holm; FDR & Benjamini–Hochberg |
13.1–13.5 |
Split lectures
Chapters 2, 3 and 4 each span two lectures. The recommended stopping points let you stop and resume cleanly:
Ch 2 — after “regression vs. classification” (p. 42); assessing accuracy, bias–variance and KNN open Lecture 2
Ch 3 — after multiple regression and the four questions (p. 76)
Ch 4 — after the logistic-regression section (p. 42); the confusion matrix, ROC/AUC and the lab open Lecture 6
One session, two decks — so it is a selection
The two precourse decks carry 157 slides in their main flow (106 + 51). A
single 180-minute session cannot cover them, and is not meant to: the session
sets up the notation, the standard-error material and the Python patterns the
chapters lean on hardest, and both decks stay available in full as the
reference. The tools for closing the rest yourself are built in — the
twelve-question self-check on page 7 of chapter_00.pdf and the notation table
on page 5 of chapter_00b.pdf — see For students.
Untaught chapters¶
Chapters 9 (Support Vector Machines), 11 (Survival Analysis) and 12 (Unsupervised Learning) are not part of the 12-lecture plan and have no lecture deck. Each ships only as a code reference notebook — Python for the methods, to be read alongside the ISLP chapter, which does the teaching — and, unlike the twelve taught labs, without worked solutions.
Assessment¶
The module is worth 6 ECTS and is graded by a single written exam at the end of the semester — 120 minutes, 100% of the mark. The Final Mock Exam is built as the rehearsal for that paper: same length, same structure, weighted to Chapters 7, 8, 10 and 13.
Practice rhythm¶
Everything below is practice, not assessment — eight papers matched to the calendar so students can self-test at the natural checkpoints, none of which counts towards the grade. See Mock exams; none is distributed with this repository.
Each deck also carries far more exercises than a session can run: the runsheets name the two to four worth live time and leave the rest as homework.
Paper |
Written after |
Covers |
Length |
|---|---|---|---|
Mock Exam 1 |
Lecture 4 |
Chapters 1–3 |
90 min · 90 pts |
Short Exam A |
Lecture 6 |
Ch 0 + 1–2, Ch 3, Ch 4 |
60 min · 60 pts |
Short Exam B |
Lecture 7 |
Ch 2, Ch 3, Ch 5 |
60 min · 60 pts |
Mock Exam 2 |
Lecture 8 |
Chapters 4–6 (+ light cumulative) |
90 min · 90 pts |
Short Exam C |
Lecture 8 |
Ch 0 + 0b, Ch 3, Ch 6 |
60 min · 60 pts |
Short Exam D |
Lecture 10 |
Ch 2 + 5, Ch 2 + 4, Ch 8 |
60 min · 60 pts |
Final Mock Exam |
Lecture 12 |
All chapters (weighted to Ch 7/8/10/13) |
120 min · 120 pts |
Short Exam E |
Lecture 12 |
Ch 0, Ch 5 + 7, Ch 13 |
60 min · 60 pts |
The three mock exams are the full-length rehearsals; the five 60-minute short exams are the formative layer, and the bold chapter is where each one’s hardest problem sits — which is why they are released in order, not all at once.
Where to go next¶
Lecture slides — the deck for each week, and how a deck is built.
Teaching it — runsheets, timings and the cut list.
Lab notebooks — the companion notebook for each chapter.
For students — prerequisites, workload and how to revise.