LEARNING OUTCOMES

The course provides algorithms and strategies to develop autonomous agents using a game engine.

AIMS AND LEARNING OUTCOMES

The aim of the course is to provide the basis for the design and development of software algorithms capable of autonomously acting within a virtual world. The student is introduced to different concepts of Reinforcement Learning and supported through extensive exercises during lectures. The course aims to train a professional figure capable of designing and implementing complex software applications using video game technologies and artificial intelligence algorithms.

PREREQUISITES

The students should have basic knowledge of programming and machine learning.

TEACHING METHODS

The course is composed of a set of frontal lessons and a set of practice sessions. During the frontal lesson, the teacher presents the topics providing also examples of live code that are tested on simualted environements. Students can use their own laptops during the lecture in order to reproduce what is proposed by the teacher. During the practice sessions, the students have to face up with real problems that they should solve by applying the techniques learnied during the lectures.

SYLLABUS/CONTENT

The titles of the main contents discussed during frontal lessons are provided in the following list. Notebooks can be obtained at the following lins: Reinforcement Learning and Deep Reinforcement Learning

01 - Markov Decision Process
02 - Dynamic Programming
03 - Exploration vs Exploitation
04 - Policy Evalutaion
05 - Policy Improvement
06 - Learning and Planning
07 - Function Approximation
08 - Deep Q-network
09 - Actor-Critic methods
10 - DDPG and PPO

RECOMMENDED READING/BIBLIOGRAPHY

• Lecture notes and notebooks (from Github)
• Books (as references):
  - Miguel Morales, Grokking Deep Reinforcement Learning, Manning
  - Richard S. Sutton, Andrew G. Barto, Reinforcement Learning: an Introduction, MIT Press