KIIS: Artificial Intelligence and Intelligent Systems 2007

A Computer Science course given in the fall semester of 2007 at Roskilde University by Henning Christiansen.

Unless otherwise stated, the course runs on Tuesdays, 9.30-12.00 and 13.00-15.30, in Room 43-2.43

Course material, notes, references, software, manuals, links to external sources such as manuals, etc. are collected in this page and linked below when relevant.

Course schedule

The schedule below will be maintained as the course goes along; when the course is finished, the "Reading" column (minus "Background reading") defines the exam curriculum.

We will work on some or all of the exercises given in the schedule for the course day; unless otherwise stated, you are expected to continue with the remaining exercises as home work. Depending on how difficult they turn out to be, we may or may not go through a solution the following course day.

Date Theme ReadingExercises, guidelines, etc.
1 4-sep-2007 Introduction to AI and the course; first Prolog workshop
Slides1, slides2 (large and strange files)
[MN] chap 1; [HC1]chapter 1 and parts of 2.
Background reading: [AT]
[HC1], exercise 2.1 and 2.2
Source texts: exercise2_1, circuits.
2 11-sep-2007 Prolog workshop continued; Rule-based expert systems; More workshop: Constraint Handling Rules in Prolog
Slides1, Slides2, Slides3.
[MN] chap 2 minus 2.7; [HC1] chap's 2 minus 2.4, chap's 3, 5 and 7.
Background reading: [WRZA]; [MN] section 2.7 (nb: example difficult to understand)
Exercises: [HC1] 5.1, 5.2 (only part), 5.3 + extra presented at lecture, 7.1 and perhaps 7.2
First written assignment. It consists of a number of smaller Prolog programming exercises, and the purpose is that you can become more familiar with Prologs way of working. You should send by electronic mail to, you answer as one pdf or word file before Monday, September 24, at 12.00.
Source texts: circuits. vertihori.txt, my_kb0Prolog.txt, expert0.txt, my_kb0.txt, gcd.txt, primes.txt, fibo.txt, kb0_as_CHR.txt.
3 18-sep-2007 Work on your own with written assigment (Your teacher is available for help Wed. 12-sep and from 20-sep and onwards) Solutions to written assignment 1: wrAssSolution1.txt, wrAssSolution2.txt, wrAssSolution3.txt.
Download these files, read them and test them until you are sure you understand them!
4 25-sep-2007 Deduction, Induction, and Abduction - with special emphasis on abduction. Application of abduction to diagnosis problem.
[HC1] chapters 7, 8 minus 8.9. Exercise about CHR.
[HC1] the exercise of section 9.2, question 1.
Source texts: db.txt, diagnosisPeriodic.txt, diagnosisConsistent.txt, deductivePower.txt, solutionCHRexercise.txt.
Solution to [HC1], 9.2 q. 1. abductivePower.txt.
5 2-oct-2007 Fuzzy expert systems [MN] chap 4; see important notes here.
Backgrund reading: [BNW] gives an alternative presentation of fuzzy logic and control.
Exercises about Fuzzy Control: Driving car using fuzzy logic; exercise 1-3, perhaps 4 and 5 if time permits (6 and 7 can be taken as project proposals beyond this course).
6 9-oct-2007 Artificial neural networks
Powerpoint slides by Angshuman Saha.
[MN] chap. 6, until and including 6.5; the rest of chap. 6 should be viewed as background material. NB: You are not expected to be able to reproduce the details of the book's formulas for adjusting weights, but you should understand the overall feed-forward, back-propagation mechanism.
We will also look at one or both of two tools for building and training neural nets which you can find at, produced by Angshuman Saha; you should read this page and the explanations provided by the two tools. NB: These tools run as Excel sheets, so they should be easy to run open and test.
Exercises: Note with exercises about Neural Networks, question 7.1-7.4; if time 7.5. Questions 7.5-7.6 recommended as homework.
Second and last part of the written assignment. You should send you answer as one pdf or word file by electronic mail to at latest Sunday, November 11.
You must give a short presentation to the other students 13-nov-2007 - or before if you are ready for it :).
7 16-oct-2007 Evolutionary computing
(We use parts the textbook's slides, lectures 9+10
[MN] chap. 7 Skip section 7.4, and be aware that there are several problematic and unclear points in this chapter; see comments.
Background reading: The following article lists interesting applications of EC/GP. J.R. Koza, M.A. Keane, M. J. Streeter: Evolving Intentions. Scientific American, Feb 2003, Vol. 288, Issue 2 (pp. unknown). Available online when you search via (works only from RUC or VPN to RUC).
Exercise for Evolutionary Computing, Questions 1 and 2
8 23-oct-2007 Natural Language processing with Definite Clause Grammars, abduction and assumptions [BBS]. (pdf document) chapters 7 and 8, minus 7.2.3 and 8.1.3; short note Natural language analysis with DCG and HYPROLOG (NB: revised 19-oct-2007); until and incl. section 5.
Background reading: [HC-CH-KT1] (pp. 40-52 of these proceedings), remaining parts of Natural language analysis with DCG and HYPROLOG, user's guide to HYPROLOG, [HC-VD1], [HC-VD2] (the last two may be a bit difficult to read.)
Exercise 1-4 of the note Natural language analysis with DCG and HYPROLOG (NB: revised 23-oct-2007)
Source texts: dcg1.txt dcg2.txt, dcg3.txt, trip.txt.
9 30-oct-2007 Statistics and Bayesian reasoning, Bayesian networks
[MN] chapter 3, until p. 62 middle (check comments in the note below:)
Examples and exercises for conditional probabilities and Bayesian reasoning (small correction 29-oct-2007).
Charniak: Bayesian Networks without Tears. AI Magazine 12(4): 50-63 (1991); skip from p. 53 "In the rest of this section, I define..." until the section headed "Consistent probabilities" p. 55. Skip also from p. 56 "Evaluation networks" and the rest of the paper.
Notice the error in fig. 2: a negation sign missing; it should read P(do | ¬ fo ¬ bd) = .3
Exercises of Examples and exercises for conditional probabilities and Bayesian reasoning; NB: Extended 29-oct-2007 with exercises for Bayesian networks.
10 6-nov-2007 More on probabilistic methods: The PRISM system, implementing Bayesian networks Course notes: Logical-Statistical models and parameter learning in the PRISM system (revised 31-oct-2007), minus section 3 (optional for those with an interest in biological sequence data).
Complementary reading (optional): Hidden Markov Models and their implementation in the PRISM system (revised 31-oct-2007).
Exercise 5.2 of Logical-Statistical models and parameter learning in the PRISM system.
Source texts: famOut.psm, famOutData.dat, hmm1.psm, hmm2.psm, hmm3.psm, words.dat.
11 13-nov-2007 Solutions to exercise 5.2 on Bayeasian networks in PRISM.
Søren Mørk on a biological application of PRISM (incl. HMM).
Students present results of written assignments.
A few words about the exam.
Two solutions to Exercise 5.2 from last week
- famOut.psm with very high prior prob's for wires being up (gives unexpected results)
- famOut.psm with prob's for wires being up depending on a weather msw (gives more natural results)
Student reports: see email!
QA 3-jan-2008
Question-and-answer session
Preparation for the exam
Please send questions or topics you want explained in advance to your teacher. If no questions or indications of interest have been received by 2-jan-2008 at 12.00, this session is cancelled.
X 10-jan-2008 EXAM Advices and guidelines for the exam
There will be a slot of total 30 min's for the exam of each student, including examiner's evaluation. Examination may take about 15-20 min's, and you are expected to start with a presentation of your written assignment, part 2, of perhaps 5 to 7 min's, which will be followed by a discussion. The examiners will most likely also ask a question to part of the course literature which is not directly related to your assignment; questions may also be asked in relation to written assignment, part 1, and possibly other parts of the course curriculum.
Those of you who did not give in written assignment part 2 cannot make a presentation, so expect more questions in the entire course curriculum.
An advice: The presentation of the assignment needs not be fancy, it's the content that matters. A possible structure could be: Which problem did I approach, how did I approach it, and what did I learn ....
Another advice: The report that you gave in for the assignment is not in itself assessed in the exam, so if you you are aware of any weak points, you need not "defend" or "repair" but you can (if you like) take these as points for discussion. If there is any doubt, contact your teacher.
Detailed time schedule is found in the exam letter that has been sent to you by email.

Last modification 21-dec-2007, Henning Christiansen