Intelligent agents

This course covers the part of Artificial Intelligence that deals with rational decision and action, as is required for intelligent agents. It starts with techniques for programming and learning optimal policies, for planning and planning with adversaries, and scaling to complex problems. The next part deals with multi-agent planning, both centralized and distributed. The third part of the course is dedicated to game theory, which provides the theoretical underpinning for multi-agent systems. 

The course is complemented with exercises in Java, using an agent platform.  The exercises form a project where students learn to implement agents and multi-agent systems of increasing complexity. For the third part of the course on game theory, paper exercises are provided.

Documents for the course are found on Moodle, and students are expected to check Moodle regularly for information. Schedules and organizational details may change in response to circumstances. The recorded lectures from 2020 (or later) are available for each week and are not very different from the live lectures. 

Grading:

There will be one midterm quiz that counts for 30% of the final grade.

There will be a tournament where each student can submit an agent and can win bonus points depending on the performance up to a maximum of 10% of the grade. The bonus points are not required to get a perfect grade in the course

There will be a final examination during the last week of the semester and it will count for 70% of the grade.

Organization:

The lectures on Wednesdays will generally last about 2 hours; the 3rd hour is dedicated to discussion of course and exercise material, and presenting exercises.

The exercise hours on Thursday are used purely to provide answer questions and provide assistance with exercises. 

Several rooms have been allocated for the exercises. However, all the assistants will initially be in BC01 and we will use the other rooms only if there is not enough space in BC01. Please come to BC01 first.

The quiz will be held on Thursday, November 7th, 13:15-14:00,  in rooms BC01, BC02, and INJ218.

The quiz is open-book: all documents are allowed. Please do not print excessive amounts of material.

The final exam will be held on Thursday, December 19th, 13:15-16:00. It is closed-book, but you are allowed one cheat sheet (A4, single-sided, two-sided if handwritten).

Please make sure you do not have a conflict with either of these dates. No special arrangements will be possible.

Introduction and Overview

Reactive Agents

Learning Agents

Deliberative Agents

Planning with Factored Representations

Deep Learning Agents

Multi-Agent Systems

Distributed Multi-Agent Systems

Nov. 7th, 13:15-14:00: Quiz

Self-interested Agents (Game-Theory)

Agent Mechanism Design

Equilibria and Negotiation

Coalitions and Group Decisions

Implementation Techniques and Applications of Agents

Final Exam: Thursday, December 19th, 13:15-16:00

Wednesday, December 18th: Q&A Session

To prepare, please note the week and slides in the course that your question refers to on the spreadsheet (link is below). Deadline: Monday, December 16th, 19:00!

I will go through the slides you reference during the QA session and you can ask you question (or put the question in the spreadsheet already).

If you have questions regarding the exercises, ask during the exercise sessions but you can mention them here as well - if enough time, I will answer them as well.

During the exam, one A4 sheet of notes is allowed - single-sided if printed, double-sided (or two sheets) if handwritten. A calculator is also allowed (although calculations are generally simple and can be done by hand).