📗

Probabilistic Machine Learning: An Introduction

Notes for each chapter


👋
1. Introduction

Links


⬇️ Download PDF
😸 Github

About these notes


How are they structured?

Each chapter is broken down into a series of question-answer pairs, with a little explanatory text sprinkled in. The point of this is not to re-write the textbook! Rather, it's a resource for students to use once they have read and understood concepts in a chapter to consolidate their knowledge over time. These notes form the basis of an Anki flashcard deck, which can be downloaded here (TODO | or on each chapter's individual page). I explain below the benefits of using a tool like Anki to aid one's retention of key ideas. The Anki decks are generated from these notion pages using the wonderful notion2anki tool. Having the notion pages here serves as a way to view the anki decks as a whole in a much friendlier UI (thanks Notion 😇) than Anki provides. It also serves as a reference and concise summary of the key ideas and definitions in each chapter. Enjoy!

Why do it this way?

A typical textbook enables a student to acquire knowledge and understanding. This is a wonderful thing, but it alone is not sufficient for the student's development. What they also need is retention - maintaining information and intuition over time, even if certain ideas are not called upon regularly. Unfortunately humans are not especially well designed for this purpose - we forget a lot, and even worse are prone to underestimating how much we tend to forget. Fortunately there are things we can do about this. One such tool is Spaced Repetition. The idea behind this approach is that the student organises the information they want to retain into small 'cards' (physically, or in software), and then uses a scheduling algorithm to decide when to test themselves. This algorithm estimates when they're on the cusp of forgetting something and tests their recall. It turns out this approach is remarkably effective! The most popular tool for this is Anki. I can attest to its effectiveness and usefullness in my own work, and there are top ML researchers also using this software. Indeed, I was inspired to start using Anki by Jeremy Howard and Michael Nielsen who both use it extensively. There are a lot of good blog posts/essays out there (like the Michael Nielsan blog I've linked) that I hope will convince you of how great tools like Anki are. Retaining knowledge in this way is not only valuable in itself, but gives a fantastic foundation for integrating new information into one's mental model of a topic. Mindless rote memorisation is rightly maligned, but ultimately some form of memorisation (and subsequent retention) is at the heart of all learning. Spaced Repetition is all about taking control over exactly what we want to retain, making memory a choice rather than an accident.