Course Readings

One reading per teaching session. Each is a self-contained text with worked, runnable code examples, a reading guide, and a before-class checklist. Session numbers follow the syllabus schedule, which has the dates and assessment calendar.

The assigned version of each reading is the PDF posted in Canvas. These pages are what the PDFs are built from, so the text is the same; the site adds working links and, where a session has them, optional deep-reading sections that are marked as such and are not part of the assigned text.

Each page opens with How to read this, which marks a core path through it. The core sections plus the before-class checklist are the required preparation, and they are the scope the module quizzes assume. Everything else — extended history, the optional worked examples, Going deeper, and the exercises — stays available as reference, but you are not expected to finish it before class.

Module 1 — Foundations

Module 2 — Convolutional Networks & Training Practice

Module 3 — Sequences, Transformers & LLMs

S13: RNNs and LSTMs
S14: Attention
S15: Guided Review (review guide for S01–S13, for self-study; S15 is not held this term)
S16: Transformer I
S17: Transformer II
S18: LLM Pretraining
S19: Fine-Tuning & LoRA
S20: Decoding, Prompting, and In-Context Learning (taught together with S21 in one class this term)
S21: Alignment: SFT, RLHF, DPO (taught together with S20 in one class this term)
S22: ViT & Multimodal

Module 4 — Generative Models, Efficiency & Evaluation

Companion resources

Demo notebooks — one Colab-ready notebook per session, containing the same worked examples as the readings.
Dive into Deep Learning and Understanding Deep Learning — free companion textbooks for the classic topics; the readings here also cover material newer than both.