course overview

The Research Seminar course is meant to study a scientific topic in a seminar format. This means that presentations are mainly held by the students themselves. Goal of the course is to learn studying, processing and presenting scientific material, and to learn about the year's topic. The seminar consists of 7 lectures, homework assignments, student presentations, a final project, a paper, and a cocktail party.

This year's topic is "Artificial Intelligence for Cocktail Parties". It covers various sexy topics from the field of Artificial Intelligence, to the level that should enable students to discuss AI comfortably at a cocktail party with other scientists. The selected topics should be of interest to Media Technology students. They were chosen to be practically applicable for future Media Technology projects. Also, they can kick-start you for further investigations into AI if you would so desire. This seminar must not be confused with the Festival for Cocktail-Robotics.

Lecturer: Maarten Lamers, of the Media Technology MSc program, LIACS institute at Leiden University
Location: room 413 of the Snellius building
Language: English
Level: 500
Credits: 4 ECTS
Requirements: experience or affinity with computer programming is recommended; a good grasp of verbal and written English is required; active participation in class is required
Enrollment: students in the Media Technology MSc program can just show up; students from other programs should send an e-mail to Maarten Lamers requesting admittance (see requirements); there is a class size restriction of 30 students
Grading: 33% homework, 33% presentation, 33% project and paper. Rounded to nearest accepted value.
Communication: information is posted on this webpage. Announcements during the course period will be posted on Mondayrunner's Media Technology Forum, students must check this forum regularly.
Schedule: included in the Media Technology MSc program calendar; also, see the detailed course schedule below

Possible follow-up courses at LIACS are:

LIACS and Media Technology staff with AI interest are:

Jacques de Vaucanson's mechanical duck

"In 1738, French engineer Jacques [de] Vaucanson built a mechanical duck that was strikingly lifelike. It could move its wings, stand up and sit down, preen itself and drink water. But what was most remarkable about it was that it seemed to be able to eat, digest and defecate -- using methods, according to [De] Vaucanson, that were copied from Nature. When the duck went on display in Paris, people flocked to see it, even with [De] Vaucanson charging an admission fee equal to a week's wages.

They wanted to see the duck not only because they had heard it was a stunning achievement of clockwork -- which, with 30 moving levers and hundreds of interlocking parts, it was -- but because they thought it might actually reflect how life worked. Even though they knew the duck wasn't really alive, they thought that you might one day be able really to produce an artificial creature."

Text quoted from Stanford Report, October 19, 2001. Drawing from Le monde des automates (1928) by Alfred Chapuis and Edouard Gélis. There are Wikipedia entries for Jacques and his digesting duck.

homework

Before classes, students must read material about the discussed topic and answer some basic test questions at the start of class. The answers are graded and the average value of all homework results makes up a student's total homework grade. If a student does not attend the lecture, a 0 is given.

A folder containing photocopies of all homework reading material is available in the office of Brit Hopmann (Snellius building, room 106). Students can borrow this folder to make photocopies of the papers only if they return the folder immediately and don't mess up the pages. Also, the corrected homework assignments are in this folder for students to view.

presentations

There are two types of student presentations:

  1. Seminar Presentations, in which two students present the homework reading material and other material that the students found themselves. Each presentation should have two parts of 20 minutes each: the first part discusses the homework reading material, the second discusses applications (or examples) of the technique and why the technique is suited for such applications. For the second part, students search find their own material.
  2. Elective Presentations, in which single students present a published scientific article of their own choice. The discussed article must be relevant to AI and of interest within this course. Elective presentations must take 20 minutes. They must start with clearly mentioning the title, authors, year of publication, and mention where it was published (just like the references to articles that I give below).
Presentation rules:

project and paper

Simply create something that applies some form of "artificial intelligence" technique to achieve a task. Choose your own AI method for this. You can build a physical thing, such as a robot or an expressive installation, or a software thing. Write a two-page paper explaining your project. You must use this Word template (also in RTF format) for writing your paper. Don't change the layout and fonts of the template -- all papers will be collected into proceedings of this course and should look alike. Project presentation: All students present their projects in class on lecture day 7. The project presentations must take 10 minutes.

cocktail party

The project presentations on lecture day 7 are concluded with a cocktail party. Actual AI researchers are invited to this also. Students can practice their AI discussion skills with them. This part of the course is not graded.

teaching assistance

Joris Slob is the teaching assistant for the Research Seminar. Joris has a solid grasp of academics, a background in the exact sciences, and much knowledge of AI. He can answer your questions regarding the homework material. If you need his help when preparing your presentation, that is fine. However, you must create the presentation yourself.
Joris is present during the lectures. Outside these hours he can be contacted via researchseminar2008@gmail.com for questions, or if you wish to make an appointment with him.

schedule

lecture datepresentationshomework reading
1 Monday January 7, 10h30 - 14h00 context and history of AI (Maarten Lamers) none
2 Monday January 14, 10h30 - 14h00 Turing Test (Nina, Bas)
artificial life (Antonis, Stijn) 		[Dennett 1990]
[Braitenberg 1984, p 1-19]
[Packard 2003]
3 Monday January 21, 10h30 - 14h00 Rodney Brooks (Antal, Lisa)
elective presentations (Lieven, Richard)
 [Brooks 1991a]
[Maes 1994]
[Linden 2003]
4 Monday January 28, 10h30 - 14h00 neural networks (Wilco, Dunya)
reinforcement learning (Karel, Klaas Jan)
 Chapter 15 of [Callan 2003]
Chapter 1 of [Sutton 1998]
5 Monday February 4, 10h30 - 14h00 evolutionary computing (Bastiaan, Casper)
affective computing (Clintonne, Stelios)
 Chapter 2 of [Eiben 2003]
[Picard 2004]
6 Monday February 11, 10h30 - 14h00 elective presentations by students
  • Eric
  • Sanne
  • Auke
  • Anne
  • Aggelos
no reading, but there will be a homework test!
7 Monday April 14, 10h30 - 16h30
+ cocktail party		 project presentations by all students,
elective presentation by Matthew,
guest lecture about DNA Computing by Hendrik-Jan Hoogeboom,
guest lecture about Artificial Compnanionship by Joost Broekens,
cocktail party with AI researchers		 none

1. context and history of AI

The course organization is discussed and an introduction to the field of Artificial Intelligence is given, by way of its relations to other fields and its history. All presentations on this day are by Maarten.

[Russell 2003] Stuart Russell and Peter Norvig (2003), Introduction, Chapter 1 (pp 1-31) of Artificial Intelligence, a Modern Approach (second edition), Prentice Hall Series in AI.
[Callan 2003] Rob Callan (2003), Introduction, Chapter 1 (pp 2-16) of Artificial Intelligence, Palgrave Macmillan.
[lecture slides] Maarten Lamers (2008), slides of introduction lecture (in PDF format) 2 slides-per-page or 6 slides-per-page.

2A. Alan Turing and the Turing Test

Alan Turing (1912-1954, UK) is the founder of computer science as we know it. His most important result was the "Turing Machine", with which he showed that all practical computing models are essentially equivalent (to put it shortly). He also played a large role within Artificial Intelligence, by developing the so-called "Turing Test".

[Turing 1950] Alan M Turing (1950), Computing Machinery and Intelligence, Mind 49 Num 236, pp 433-460. (only accessible from the University's computer network)
[Dennett 1990] Daniel C Dennett (1990), Can Machines Think?, from The Age of Intelligent Machines, Ray Kurzweil, MIT Press, pp 48-61.
[Dennett 1997] Daniel C Dennett (1997), Consciousness in Human and Robot Minds, from Cognition, Computation, and Consciousness, by M Ito, Y Miyashita, and ET Rolls, Oxford University Press.
[Copeland 2000] B Jack Copeland (2000), The Turing Test, Minds and Machines 10, pp 519-539.
[Wikipedia] Wikipedia Entry for Alan Turing.

2B. artificial life and emergence

The field of Artificial Life attempts to simulate and study processes that we associate with living. It is a broad, and somewhat ill-defined field, but interesting nonetheless.

[Braitenberg 1984] Valentino Braitenberg (1984), Vehicles: Experiments in Synthetic Psychology, MIT Press, pp 1-19.
[Conway 1970] Wikipedia Entry for Conway's Game of Life.
[Steels 1994] Luc Steels (1994), The Artificial Life Roots of Artificial Intelligence, Artificial Life Journal, Vol 1 Num 1-2, pp 75-110.
[Sipper 1995] Moshe Sipper (1995), An Introduction to Artificial Life, Explorations in Artificial Life (special issue of AI Expert), pp 4-8.
[Packard 2003] Norman H Packard and Mark A Bedau (2003), Artificial Life, Encyclopedia of Cognitive Science, Vol 1, Macmillan Publ., pp 209-215.
[Velleman Microbug] The Velleman company created a cheap (~10 EURO) Braitenberg vehicle kit, called the MK127 Microbug. Some electronics supplies stores sell it, or you can order it at many online electronics stores.

3A. Rodney Brooks and the subsumption architecture

Rodney Brooks was the director of MIT's Computer Science and Artificial Intelligence Laboratory from 1997-2007. He holds highly interesting ideas about AI. Resulting from these ideas, he developed the "subsumption architecture", an abstract architecture for incrementally building autonomous robots.

[Brooks homepage] Rodney Brooks' homepage.
[Brooks 1985] Rodney Brooks (1985), A Robust Layered Control System for a Mobile Robot, MIT AI-Lab Memo 864.
[Brooks 1990a] Rodney Brooks (1990), Elephants Don't Play Chess, Robotics and Autonomous Systems 6, pp 3-15.
[Brooks 1990b] Rodney Brooks (1990), Artificial Life and Real Robots, in Toward a Practice of Autonomous Systems: Proceedings of the First European Conference on Artificial Life, MIT Press, pp 3—10.
[Brooks 1991a] Rodney Brooks (1991), Intelligence Without Representation, Artificial Intelligence 47, pp 139-159.
[Brooks 1991b] Rodney Brooks (1991), Intelligence Without Reason, Proceedings of 12th Int Joint Conf on Artificial Intelligence (Sydney), pp 569–595.
[Brooks 1991c] Rodney Brooks (1991), New Approaches to Robotics, Science 253, pp 1227-1232.
[Brooks 2002] Rodney Brooks (2002), "De Kunstmatige Mens. Hoe Machines Ons Veranderen" is on sale very cheap at De Slegte bookstores. The book's original title is "Flesh and Machines. How Robots Will Change Us".
[Rekveld] Joost Rekveld's Autonomous Robots webpage contains other interesting information.

3B. collaborative filtering

Collaborative filtering is a method of making automatic predictions about the interests of a user by collecting taste information from many other users. The underlying assumption is that those who agreed in the past tend to agree again in the future.

[Maes 1994] Pattie Maes (1994), Agents That Reduce Work and Information Overload, Communications of the ACM 37(7), pp 30-40. (only accessible from the University's computer network)
[Shardanand 1995] Upendra Shardanand and Pattie Maes (1995), Social Information Filtering: Algorithms for Automating "Word of Mouth", Proceedings of ACM SIGCHI'95.
[Linden 2003] Greg Linden, Brent Smith, and Jeremy York (2003), Amazon.com Recommendations: Item-to-Item Collaborative Filtering, IEEE Internet Computing Vol 7(1), pp 76-80.

4A. feedforward neural networks

Feedforward networks are a specific class of artificial neural networks. Basically, these are models that can model relationships between pairs of data. For example, between the age, sex, background of a student and that student's grade in the Research Seminar. By applying a specific algorithm (e.g. backpropagation) they can adapt their internal parameters until they model relations in a dataset. This is commonly done by repeatedly presenting the network with example input/output pairs of data.

[Weiss 1991] Weiss and Kulikowski (1991), Sections 4.1 - 4.3 (pp 81-102) of Computer Systems That Learn: Classification and Prediction Methods from Statistics, Neural Nets, Machine Learning, and Expert Systems, Morgan Kaufmann.
[Callan 2003] Rob Callan (2003), Neural Networks I, Chapter 15 (pp 286-311) of Artificial Intelligence, Palgrave Macmillan.
[Russell 2003] Stuart Russell and Peter Norvig (2003), Neural Networks, Section 20.5 (pp 736-748) of Artificial Intelligence, a Modern Approach (second edition), Prentice Hall Series in AI.

4B. reinforcement learning

Reinforcement learning is a field of artificial intelligence wherby an agent tries to maximize the total reward it receives when interacting with a complex, uncertain environment.

[Kaelbling 1996] Leslie P Kaelbling, Michael L Littman, and Andrew W Moore (1996), Reinforcement Learning: A Survey, Journal of Artificial Intelligence Research Vol 4, pp 237-285.
[Sutton 1998] Richard S Sutton and Andrew G Barto (1998), Introduction, Chapter 1 of Reinforcement Learning: An Introduction , MIT Press.

5A. evolutionary computing

Evolutionary computing is the collective name for a range of problem-solving techniques (algorithms) based on principles of biological evolution, such as natural selection and genetic inheritance. Basically, their strategy is to create a collection of possible solutions for your problem (which may be very bad solutions) and to use the principles of evolution to evolve better solutions from this collection.

[Eiben 2003] AE Eiben and JE Smith (2003), Introduction to Evolutionary Computing, Springer.
- Chapter 1, Introduction, pp 1-14
- Chapter 2, What is an Evolutionary Algorithm?, pp 15-35
- explanation of symbols used in Chapter 2,
- Guszti Eiben worked at LIACS before becoming a professor in Amsterdam.
[De Jong 2006] Kenneth A De Jong (2006), Introduction, Chapter 1 (pp 1-22) of Evolutionary Computing, a Unified Approach, MIT Press.

5B. affective computing

Affective computing is a field within artificial intelligence that deals with devices that can process emotions. It is an interdisciplinary field, relating to computer science, psychology, and cognitive science.

[Picard 1997] Rosalind W Picard (1997), Affective Computing, MIT Press.
[Picard 2000] Rosalind W Picard (2000), Toward Computers That Recognize and Respond to User Emotion, IBM Systems Journal, Vol 39 Num 3-4, pp 705-719.
[Picard 2004] Rosalind W Picard, et al. (2004), Affective Learning — a Manifesto, BT Technical Journal, Vol 22(4), pp 253-269.

6. elective presentations

On this day, students who did not participate in a seminar presentation must give an elective presentation about a published scientific article of their choice. The lecture ends with a class discussion about artificial intelligence.

7. project presentations

On this day, students present their projects and hand in their papers. The day ends with a cocktail party to which AI researchers are also invited.