University of Bologna · Prof. Matteo Ferrara · a.y. 2025/2026

Deep Learning

12 chapters54 interactive widgets63 plates~9 hours of study

Plate 00 — The road of the course. Foundations (math and automatic differentiation), then neural networks and how to train them, then the architectures that made deep learning scale (CNNs, RNNs, autoencoders, transformers), and finally the two frontiers: generating data and learning from interaction.

How to study

The chapters are in dependency order: read them from the beginning. Every chapter consolidates in one place everything the lectures said about a topic.
The plates are archival technical diagrams reconstructed from the slides: read them as figures, not decoration. The widgets (simulators, steppers, explorers) are worked implementations of the mechanisms — use them actively.
Each chapter ends with “Check your understanding”: collapsible exam-style questions with answers in the exact wording of the course.
Every chapter footer lists the lecture deck it comes from; sources are the official course slides, not redistributed.

Foundations

Tav. I — Forward path, backward gradient

Neural networks and training

Tav. II — Descent into the minimum

III

Architectures for structured data

Tav. III — Grids, sequences, attention

Generative models

11Generative models55 min · 4 widgets

Tav. IV — The adversarial game and the denoising chain

Reinforcement learning

12Reinforcement learning55 min · 4 widgets

Tav. V — The agent–environment loop