Machine Intelligence
70 PAGES · 7 TOPICS · 0 DONE OF 62Search, genetic algorithms, trees, nets, Bayes and clustering - every algorithm derived, then drawn.
Machine Intelligence
A course in how machines learn, filed by topic rather than by lecture. Every page answers one question: what is the algorithm actually doing, and why does that work? Derivations are kept, hand-waving is not.
Read it start to finish, or jump to a topic from the index above. Each topic closes with a quick-revision cheat sheet built for the night before the exam.
How to read a page
Each page is built from labelled cards, and the label on a card tells you what kind of thing is inside it before you read a word of it.
- Facts - named terms with short definitions, the definitional lists.
- Steps - a sequence where the order is the content, including every worked numeric example.
- Numbers - the constants and thresholds worth memorising.
- Compare - two or three named things set side by side, with no rule for picking.
- Decision - a question with a
pick this whenanswer. - Exam cue - one-directional pattern match: if the question says X, the answer is Y.
- Best practice - one recommendation and the reason for it.
- Gotcha - a trap that will cost you marks. Where the source slides contradict themselves or round a number wrong, a gotcha says so rather than quietly correcting them.
Where each unit went
The notes are filed by topic, not by lecture, so one unit's sections are spread across several topics. If you need to revise a unit end to end:
| Unit | Sections | Topics that carry it |
|---|---|---|
| Unit 1 | §1-9 | Foundations & Evaluation, Search & Optimization, Supervised Models, Learning Theory |
| Unit 2 | §1-12 | Supervised Models |
| Unit 3 | §1-13 | Ensemble Learning, Probabilistic & Bayesian Learning |
| Unit 4 | §1-10 | Unsupervised Learning |
| Unit 5 | §1-11 | Search & Optimization, Learning Theory |
Every page carries its own Source: Unit N §M line under the heading.
Start with Machine Learning Foundations.