- Knowledge of the main characteristics of artificial intelligence: historical overview and modern definition - Knowledge of how to formalize a learning problem from the mathematical foundations of the optimization objectives to the composition of a deep architecture - Knowledge of the main characteristics of modern deep learning techniques with practical engineering tricks for end-to-end training and fine-tuning the networks - Basic knowledge about how to deal with research questions, how to organize a project and prepare a scientific conference report

- Linear Algebra - Probability theory concepts - Basic concepts of decision theory (model optimization) - Python: Basic elements

Module 1 - Introduction (6h Theory, 3h Laboratory): - Artificial Intelligence: historical definition, brief overview, modern definition and current role of machine and deep learning - Overview of fundamental knowledge of probability - Overview of decision theory: loss, risk, Probably Approximately Correct (PAC) Learning - Python concepts, brief recap Module 2 - Neural networks basics (18h Theory, 9h Laboratory): - Perceptron and Artificial Neural Networks - Convolutional Neural Networks basic algorithms: Backpropagation and Stochastic Gradient Descent - Training a CNN (data preprocessing, weight initialization and hyperparameter optimization) - Visualizing and Understanding the CNN inner working - Classification and detection applications Module 3 - Advanced learning topics (16h Theory, 8h Laboratory): - Multi-Task and Structured Output Learning (Semantic Segmentation and Detection) - Unsupervised and Metric Learning (Siamese networks and Contrastive Learning for Retrieval) - Recurrent Neural Networks - Learning with few samples and across domains - Generative Networks (Autoencoders and GANs) - Transformers - Applications (for example, video understanding, graph neural networks, spatial reasoning) Besides the lectures, the course will include selected topic-dedicated reading group sessions where the students will actively participate by presenting and discussing recent research papers.

- 40 hours of lectures (theory section) - 20 hours of 'laboratory' exercises (practical section)

Slides + State of the art deep learning top conference and journal papers. Optional: specific pointers to chapters of the following books will be indicated during the lectures. - Artificial Intelligence: A Modern Approach S. Russell, P.Norvig - Deep Learning by I. Goodfellow, Y. Bengio, A. Courville. - Dive into Deep Learning by A. Zhang, Z. C. Lipton, M. Li, A. J. Smola, et. al. - Understanding Machine Learning: From Theory to Algorithms by Shai Shalev-Shwartz and Shai Ben-David - Machine Learning: a Probabilistic Perspective by Kevin P. Murphy - The Elements of Statistical Learning, T. Hastie, R. Tibshirani, and J. Friedman - Pattern Recognition and Machine Learning, Christopher M. Bishop - Machine Learning, Tom M. Mitchell - A Course in Machine Learning, Hal Daumé III - Neural Networks - a Systematic Introduction, Raul Rojas

Video lectures (current year);

Exam: Compulsory oral exam; Group project;

- (30%) Paper presentation during the course: the students are asked to read, present, and discuss a recent research paper. The goal of this part of the exam is to evaluate the students' comprehension of a scientific research text, by identifying: (a) what is the task that the authors are studying; (b) how the newly proposed approach relates to previous work; (c) the intuition at the basis of the proposed algorithm; (d) the experimental setting and metrics; (e) discuss the results and explain potential limitations. The presentation is usually performed in groups (3 students), followed by individual questions. - (40%) Project Work. It consists in reproducing and extending a current state-of-the-art deep learning method. The students can choose among a set of topics and are asked to re-implement an existing approach, run experiments, and test some suggested extensions. The students will work in groups and submit 10 days before the oral exam a PDF report of 8 pages in IEEE paper format. This part of the exam aims at evaluating the students' ability in coding and experimentally validating a deep learning model as well as reporting and discussing the obtained results. The evaluation process includes a check on the code and a review of the submitted pdf report. Extra points (for honors, "lode") are assigned to original, well-justified, and experimentally validated method extensions proposed by the students. - (30%) Oral Presentation and Theory questions. This part of the exam aims at verifying the equal contribution of all the authors to the project work. The students are asked to present the work done. The content and timing of the presentation will be evaluated as well as the answers to individual questions. The students are also asked about more general theory topics presented in class.

In addition to the message sent by the online system, students with disabilities or Specific Learning Disorders (SLD) are invited to directly inform the professor in charge of the course about the special arrangements for the exam that have been agreed with the Special Needs Unit. The professor has to be informed at least one week before the beginning of the examination session in order to provide students with the most suitable arrangements for each specific type of exam.