Machine Learning Summer School 2026

Welcome

Slides, reading companion chapters, notebooks, and a short map of the week — all in one place.

Use this site during and after class: open today’s deck, then try the matching notebook or lab exercise when you feel like coding along. If something is unclear, that’s normal — bring it to the room or office hours.

Tip Start here (2 minutes)
  1. Open today’s slides — we follow Day 1 (Monday) → Day 2 (Tuesday) → Day 4 (Thursday) in class.
  2. When you’re ready, open a reading companion chapter for explanation, or a notebook for a long IDE lab — same R you already use (RStudio or Positron is fine).
  3. Skim “Your week” below for the story of the three days — no jargon quiz attached.

Palmer Penguins

Our real-data running example

  • Data catalog — data cards for every dataset (penguins, genes, Old Faithful, Hald, microbiome)
  • Palmer Penguins card — variables, biology, caveats, teaching slices Monday → Thursday
  • Links to penguin-only notebooks from that page (mass, binary species, sex, three-class species); tidymodels walk-throughs live in Chapter 4 (+ 7–8 follow-ups; see the reading companion).

Synthetic gene notebooks are there so we know the true story behind the data (which “genes” really matter) while we teach LASSO, ridge, and AIC.

Class slides

Follow along day by day

  • Day 1 (Monday) — from questions to simple models: when a straight line (or curve) helps, and when a model is “trying too hard.”
  • Day 2 (Tuesday) — branching decisions (trees), and how to keep your code organized so Day 4 still makes sense.
  • Day 4 (Thursday) — forests, boosting, neural nets, PCA/imputation in recipes, variable importance, and evaluation.

Each deck links back here from the footer if you get lost.

Printable PDF: open a deck, click Print PDF in the footer (or add ?print-pdf to the URL), then Save as PDF in Chrome (Landscape, no margins, background graphics on).

Hands-on notebooks

Try the code yourself

Go as fast or slow as you like; notebooks and chapters are for you, not for a grade on this site.

Lab exercises

Afternoon practicals (with solutions)

  • Lab exercises home — task handbook plus links to every solution notebook (HTML + source)
  • Lab exercises — microbiome afternoon tasks (Day 1, 2, 4) with lecture placement hints
  • No grade on this site: try the tasks yourself first; solutions are there when you need them.

Outline & reading companion

If you want more structure

  • Curriculum outline — how the three days connect (also useful if you miss a session)
  • Reading companion — eight chapter pages that explain the ideas in prose, with light code and links back to slides, labs, and notebooks.

Your week at a glance

Think of it as one story in three chapters: see the pattern, keep your work honest, then choose and explain models without pretending they’re magic. Curiosity beats perfection.

  • Day 1 (Monday) — predictions, when a model is “too wiggly,” and gentle introductions to linear models, logistic regression, and a bit of regularization. You’ll see simulated genes (we scripted the biology so we know what’s true) and Palmer Penguins (real measurements) side by side.
  • Day 2 (Tuesday) — decision trees (“if bill length is above this line, go left…”) and how to fit and check models in a tidy, repeatable way in R so Day 4 still makes sense.
  • Day 3 (Wednesday) — mini symposium and social event (no lecture deck on this site).
  • Day 4 (Thursday) — stronger models, what to watch when outcomes aren’t balanced, and how to discuss why a model leaned one way — without claiming it “knows” biology or medicine by itself.

Organizers & people

This summer school is hosted by the Interfaculty Bioinformatics Unit (IBU) at the University of Bern.

Host institute

Organizers

Lecturers

Teaching assistants

Funding

  • CUSO — Conférence universitaire de Suisse occidentale (doctoral programme)

CUSO SIB Swiss Institute of Bioinformatics University of Bern

  • Local preview: from this project folder run quarto preview so links resolve; after quarto render, you can serve _site/ with any static file server.

  • Slide decks (Reveal.js): use the navbar Day links or open _site/slides/day-02-tuesday.html after render. IDE preview of a .qmd file shows a scrolling article — that is not the slide deck. To preview one deck: quarto preview slides/day-02-tuesday.html.

  • Course design detail — the curriculum outline describes the two-track story (genes + Palmer Penguins), calibration, causal guardrails around explainability, data/model cards, reproducibility, and assessment prompts.

  • Site output lives in _site/; GitHub Actions builds the same way for Pages.

  • Palmer Penguins — Data card and full catalog; Day 2 (Tuesday) modeling pipeline examples use the same Adelie vs Gentoo setup as the penguin classification notebook (implemented with the tidymodels R packages).

  • Single source for the canonical penguin pipeline — day02-tidymodels-walkthrough.qmd (Day 2 slides + Chapter 4 via tidymodels-pipeline.qmd wrapper).

  • Slide reveal mode — edit incremental in slides/_reveal-config.yml (false = full slide at once; true = step-by-step bullets), then re-render the day deck.

  • Reading chapter (L1/L2, lasso, elastic net) — Chapter 2.