Quick start

You don’t need to install anything to read the slides — the compiled PDFs are linked from Lecture slides. To run a lab, start in Colab. Install locally later, in your own time, once you know you want to.

▶︎ Day one: Google Colab — nothing to install

This is the documented first-session route, and it is the one to use in the room. Open any notebook in your browser; there is nothing to set up and nothing to go wrong on a projector.

Every notebook’s first cell detects Colab, quietly installs the few missing packages (ISLP, plus pygam / xgboost / lifelines where a chapter needs them; torch is preinstalled on Colab), and resolves the data automatically — datasets load straight from the ISLP package wherever possible, and the four the package does not ship (Advertising, Heart, Income1, Income2) stream from the book’s official site.

One-click links for all fifteen notebooks are on the Lab notebooks page.

Tip

The Colab links open straight from the public GitHub repository — you only need a Google account to run a notebook, not any access to this repository.

Colab is enough for every lab in the course, including the Chapter 10 deep learning lab.

⌥ Week two: a local virtual environment

Worth doing once the course is under way, and worth doing properly: a local environment is faster, works offline, keeps your edits, and is what you will want for any serious piece of work. It is not a first-session activity — the install pulls in around 150 packages and several hundred megabytes, because the book companion package ISLP hard-requires torch (together with pytorch_lightning and torchmetrics). On Windows torch alone is over 100 MB; on Linux it is several times that, as the wheel bundles the CUDA libraries. Budget time for it, on a decent connection, outside class.

python -m venv .venv
source .venv/bin/activate         # Windows: .venv\Scripts\activate
pip install -r requirements.txt
jupyter lab Chapters/chapter_03/chapter_03_lab.ipynb

Tested with Python 3.9+. Data loads via the ISLP package when installed, with an automatic fallback to the bundled ALL CSV FILES - 2nd Edition/ folder — see Datasets, and Python environment for what is pinned and why.

Rebuilding the LaTeX materials

The decks are compiled from LaTeX sources kept in the repository, so anything can be edited and rebuilt. This needs a TeX Live distribution with beamer, tcolorbox, tikz, listings and booktabs.

cd Chapters/chapter_03
pdflatex chapter_03.tex
pdflatex chapter_03.tex   # second pass for the navigation bar

pip install -r docs/requirements.txt
sphinx-build -b html docs docs/_build/html
open docs/_build/html/index.html

See Building the docs for live-reload and PDF output.

Where to go next