- Docente: Elena Bandini
- Credits: 6
- SSD: MATH-03/B
- Language: Italian
- Teaching Mode: In-person learning (entirely or partially)
- Campus: Bologna
- Corso: Second cycle degree programme (LM) in Statistics, Economics and Business (cod. 6811)
-
from Sep 15, 2026 to Oct 22, 2026
Learning outcomes
The course aims to provide students with advanced knowledge of probability and inference, with particular reference to sequences of random variables and convergence theorems, as well as to develop their knowledge, skills, and competencies in reinforcement learning methods based on Markov decision processes.
The course includes laboratory applications using Python software.
By the end of the course, students will be able to apply reinforcement learning methods to model and solve data-driven decision-making problems.
Course contents
Discrete-time Markov chains.
Markov decision processes.
Dynamic programming algorithms: policy evaluation, policy improvement, policy iteration, value iteration.
Model-free algorithms: Monte Carlo prediction, temporal-difference prediction, TD(lambda), Monte Carlo control, temporal-difference control, SARSA, Q-learning.
Readings/Bibliography
Slides provided by the instructor, available on the virtuale.unibo.it platform.
Supplementary textbook:
R. S. Sutton and A. G. Barto, Reinforcement Learning: An Introduction, Second Edition, 2018.
Teaching methods
Theoretical lectures will alternate with examples of applications using pseudocode and Python code.
Assessment methods
There will be a written exam lasting approximately two hours, consisting of three exercises covering the topics addressed in the course.
Teaching tools
The course website is available on the virtuale.unibo.it platform, where students can find the course slides, worked exercises, and other useful information about the course.
Office hours
See the website of Elena Bandini