42026 - Machine Learning in Agents and Multi-agent Systems (AASM) [UPC]

The aim of this course is to study the main characteristics of Multiagent Systems in Artificial Intelligence. The two most important issues that we deal with are Learning Behaviours as the adaptive component of agents and impact that interactions between agents have on learning.

Learning in agents differs a lot from other types of Machine Learning already studied as the aim is learn behaviour to follow when the task is to accomplish a goal instead of a data classification.

The material of this course complements “Multiagent Systems".



1. Differential aspects in Learning Behaviours
2. Reinforcement Learning (RL)

• RL Framework
• Reinforcement functions
• Evaluation functions

3. Optimization Policies

• Evaluation functions and optimal policies
• Dynamic Programming
• Reinforcement Learning Algorithms

4. Learning Behaviours under uncetainty

• Sources of uncertainty
• POMDPs Framework
• Solutions to the POMDPs

5. Generalization of Learning Behaviours
6. Knowledge sharing between processes


1. Knowledge about other agents
2. Learning Algorithms is Multiagent Systems
3. Applications oflearning to Multiagent Systems