This course provides an introduction to reinforcement learning methods for control, beginning with an overview of the fundamental concepts of machine learning. Particular emphasis will be placed on neural networks and deep learning techniques, which constitute the core of many modern reinforcement learning approaches. Building on these foundations, the course will introduce the main principles and algorithms of reinforcement learning, with a focus on their application to control and decision-making problems.

- knowledge of the fundamental concepts of machine learning and reinforcement learning; - knowledge of neural networks and deep learning techniques relevant to reinforcement learning applications; - learn to model control and decision-making problems within the reinforcement learning framework; - understand the principles of agent-environment interaction, reward design, and policy optimization; - develop the ability to apply reinforcement learning techniques to simple control problems and interpret the obtained results

- Linear Algebra - Basic concepts of probability theory - Python

The course is subdivided in two main modules: Part 1 -- Introduction to Machine Learning (approx. 20h) - Overview of machine learning paradigms (supervised, unsupervised, and reinforcement learning) - General concepts of machine learning (data representation, loss functions, dataset partitioning) - Fundamentals of neural networks - Convolutional Neural Networks basic algorithms (backpropagation, Stochastic Gradient Descent) - Overview of deep learning architectures - Training a neural network (data preprocessing, weight initialization and hyperparameter optimization) Part 2 --Introduction to Reinforcement Learning (approx. 40h) - Basic concepts of reinforcement learning (the reinforcement learning problem: state, actions, rewards, and policies) - Multi-arm Bandits - Markov Decision Processes - Dynamic programming - Model-free prediction (Monte-Carlo methods, temporal difference learning) - Value function approximation - Policy gradient methods - Model-based reinforcement learning

40h lectures + 20h hands-on labs

The main learning resources are slides and supplementary material provided by the teachers. Supplementary material: - Reinforcement Learning: An Introduction, Richard S. Sutton, Andrew G. Barto

Slides; Esercizi risolti; Esercitazioni di laboratorio;

Modalita di esame: Prova orale obbligatoria;

Exam: Compulsory oral exam;

... The exam aims at assessing students' knowledge and understanding of the topics covered in the course, as well as their ability to critically discuss such topics. The exam is oral, and students are required to answer questions covering the entire course content. Students are not allowed to use neither books nor lecture notes during the exam. The oral exam is divided in two parts. The first part focuses on the theoretical aspects of the course and contributes up to 24 points to the final score. The second part focuses on the experiments conducted during the in-class labs and contributes up to 8 points. The maximum achievable score is 32. Honors (lode) will be awarded to students who achieve a final score of 31 points or higher.

Gli studenti e le studentesse con disabilita o con Disturbi Specifici di Apprendimento (DSA), oltre alla segnalazione tramite procedura informatizzata, sono invitati a comunicare anche direttamente al/la docente titolare dell'insegnamento, con un preavviso non inferiore ad una settimana dall'avvio della sessione d'esame, gli strumenti compensativi concordati con l'Unita Special Needs, al fine di permettere al/la docente la declinazione piu idonea in riferimento alla specifica tipologia di esame.