Neural Networks for Data Science Applications
Master's Degree in Data Science (2024-2025)
For the previous year (2023-2024), refer to this page.
Important Info
| Material: slides, assignments, and grading will be done via Google Classroom. | |
| Timetable: Wednesday 5-7 PM (Aule A5 + streaming in A6, Via Ariosto), Friday 8-11 (Aule A5 + streaming in A6, Via Ariosto). |
News
- Classes will start on September 23rd (see the faculty calendar).
General overview
The course provides a general overview on neural networks as compositions of differentiable blocks, that are optimized numerically. We describe common building blocks including convolutions, self-attention, batch normalization, etc., with a focus on image, audio, and graph domains. The course combines rigorous mathematical descriptions with many coding sessions in TensorFlow.
More information about the topics, the exam, organization, etc. can be found on the introductory slides.
Teaching assistants
Material
Lab sessions implemented in JAX are in blue. Homeworks and projects (mandatory) are in red. Seminars (optional) are in green.
| Â | Date | Content | Material |
|---|---|---|---|
| L0 | 25/09 | About the course | Slides |
| L1 | 25/09 | Introduction | Slides |
| L2 | 27/09 | Preliminaries | Slides Video (1/2) Video (2/2) |
| L3 | 02/10, 04/10 | Supervised learning | Slides Video (1/2) Video (2/2) |
| - | - | The deep learning software ecosystem | Slides |
| E1 | 09/10, 11/10 | Lab 1: logistic regression in JAX | Notebook |
| L4 | 16/10 | Fully-connected models | Slides Video (1/2) Video (2/2) |
| L5 | 27/11 | Automatic differentiation | Slides |
| L6 | 23/10 | Convolutional layers | Slides Video (1/2) Video (2/2) |
| Â | 30/10 | Lab 2: CNNs in Keras | Notebook |
| - | 05/11 | Interpretability for Language Models: Current Trends and Applications (Gabriele Sarti, University of Groningen) | Webpage |
| - | 07/11 | Towards interpretability-by-design (Pietro Barbiero, USI) | Webpage |
| L7 | 06/11 | Convolutions beyond images | Slides Video |
| - | 11/11 | A Journey into PyTorch, the Ecosystem, and Deep Learning Compilers (Luca Antiga, CTO, Lightning AI) | Webpage |
| H1 | 13/11 | Mid-term homework: custom activation functions | Template Video instructions |
| L8 | 13/11 | Building deeper models | Slides Video (1/2) Video (2/2) |
| E3 | 15/11 | Lab 3: neural networks in Equinox | Notebook 1 Notebook 2 |
| L9 | 29/11 | Transformer models | Slides Video (1/3) Video (2/3) Video (3/3) |
| L10 | 11/12 | Recurrent models | Slides Video (1/2) Video (2/2) |
| H2 | - | End-of-term homework: recurrent models in JAX | Template Video |
Book: Alice’s Adventures in a Differentiable Wonderland
The course is complemented by a book which expands on most topics covered during the lectures:
- Buy the book on Amazon (independently published to keep the price low).
- Downloaded the updated full draft (08/07/2024).
For the full book webpage (with arXiv version and errata list): https://sscardapane.it/alice-book/