Deep Learning & Applied AI @Sapienza
Course material, 2nd semester a.y. 2020/2021, Dept. of Computer Science
- 14/09/2021: The scores for the written exam (September 14) are now published; please refer to this issue for details.
- 28/04/2021: The list of projects has been published. Please refer to this issue for further details.
- 21/04/2021: Please fill out the OPIS questionnaire (instructions here). The code for this course is KIW0A907.
Lecturer: Prof. Emanuele Rodolà
Assistants: Dr. Luca Moschella, Dr. Antonio Norelli
When: Wednesdays 09:00--12:00 and Thursdays 17:00--19:00 (official schedule)
- Physically: Aula 1 - Aule L Via del Castro Laurenziano 7a
Important: Please use Infostud Lab (not Prodigit!) for booking a seat in the classroom. This should be done no later than 48h from the start of the lecture.
- Virtually: Zoom (link to the virtual classroom), Meeting ID: 475 234 9941, Passcode: 3K7xrM.
Q & A: Please use the issue system of Github. Here is the link to the course repository, you'll need to create a free account to access it.
Programming fundamentals in Python; calculus; linear algebra.
Textbook and reading material
Due to the continuously evolving nature of the topic, there is no fixed textbook as a reference. Specific material in the form of scientific articles and book chapters will be given throughout the lectures.
Evaluation proceeds according to the following steps:
- A midterm self-evaluation test (optional, does not concur to the final grade)
- A final written exam (mandatory, accounts for 60% of the final grade)
- A project (mandatory, accounts for 40% of the final grade)
- An oral exam (optional, attributes at most 3 points, added to or subtracted from the final grade)
The cum laude can be obtained only by taking the oral exam. For students who already have a very high score with written exam + project, the oral exam is meant to confirm the high score.
The list of projects is now published (details here). Each project must be accompanied with code to a github repository, and with a 2 page report using this fixed template. Projects can be made in groups of at most 2 students.
Here you can find exam sheets for the written part from past exam sessions:
All the lectures (video, audio and chat script) will be recorded and stored in this Google Drive folder.
Please note that the audio quality will not be very good in general, due to technical issues.
|Date||Topic||Reading||Code & Data|
|Wed 24 Feb||Introduction||slides|
|Thu 25 Feb||Data, features, and embeddings||slides|
|Wed 03 Mar||Tensor manipulation|
|Thu 04 Mar||Linear algebra revisited||slides|
|Wed 10 Mar||Tensor operations|
|Thu 11 Mar||Linear regression, convexity, and gradients||slides; notes on matrix meta-mechanics|
|Wed 17 Mar||Linear models and Pytorch Datasets|
|Thu 18 Mar||Going nonlinear, overfitting, and regularization||slides|
|Wed 24 Mar||Logistic Regression and Optimization|
|Thu 25 Mar||Stochastic gradient descent||slides|
|Wed 31 Mar||Multi-layer perceptron and back-propagation||slides|
|No lecture due to Easter Holidays|
|Wed 07 Apr||Autograd and Modules|
|Thu 08 Apr||Midterm self-evaluation test||midterm; evaluations|
|Wed 14 Apr||Convolutional Neural Networks|
|Thu 15 Apr||Convolutional Neural Networks||slides|
|Wed 21 Apr||Uncertainty, regularization and the deep learning toolset|
|Thu 22 Apr||Invited lecture by Giambattista Parascandolo Memorization and Invariances in Neural networks & Learning abstract models||slides||Learning explanations that are hard to vary; Adaptive skip intervals; Teacher-student framework|
|Wed 28 Apr||Variational AutoEncoders|
|Thu 29 Apr||Regularization||slides|
|Wed 05 May||Deep generative models||slides|
|Thu 06 May||Adversarial training||slides|
|Wed 12 May||CycleGAN and Adversarial Attacks|
|Thu 13 May||Geometric deep learning||slides; video by Michael Bronstein|
|Wed 19 May||Self-attention and transformers||slides|