[Sydney Uni header]
School of Information Technologies

COMP 3002/ COMP3902 
Artificial Intelligence

Lecturer:  Irena Koprinska


Course Outline
Timetable
Resources
Assessment Overview
Assignments    Exam
Message Board
Lecture Notes
Tutorial Exercises
Marks

News

[30/5/03] A sample exam paper is up. The solutions will be discussed at lecture 2, week 12 and tutorials in weeks 12 and 13.
[30/5/03] The marks have been updated - all (h1, h2 and a1) should be there except for h2 of Harry's class. h3 will be posted by the end of week w12 and ass2 by the end of week 13. Please check your marks regularly and report  problems to your tutor.
[21/5/03] Please check your marks. Some marks are still missing, will be added as they become available.
[16/5/03] Prof. Patrick's lectures on natural language processing have been moved to week 13.
[16/5/03] Clarifications re assignment 2 (should you need them):
    -For the implementation you can use a high-level language of your choice except Matlab's Neural Network toolbox (or other similar toolboxes where the algorithm is already implemented ). However, Matlab can be used for the other parts of the assignment, e.g. the cross validation.
    -You have to implement 10-fold cross validation (in addition to the backpropagation algorithm).
[5/5/03] Assignment 2 is released.
[27/4/03] Assignment 2 will be released by the end of this week. It will be on backpropagation neural networks and the material is not covered yet.
                Week 6  lecture (decision trees) - additional slides (with answers and more details on the information gain heuristic) are provided.
[13/4/03] Reminder: if your tutorial class is on Friday, assignment 1 is due on Thursday midday, 17 April. Please submit it in the locker of your tutor (John Drake or Clinton Freeman). Lockers are located in Madsen Building, left coridor.
[28/3/2003] Assignment 1 is released.
[25/3/2003]  The deadline for assignment 1 has been extended to week 6 (as we still haven't covered how to create admissible heuristics that is needed for the assignment). The assignment will be released Thursday this week

Outline

Welcome to COMP3002/3902 Artificial Intelligence!

The objectives of this course are to provide an introduction to some Artificial Intelligence techniques. Artificial Intelligence is all about programming computers to perform tasks normally associated with intelligent behaviour. Classical AI programs have played games, proved theorems, discovered patterns in data, planned complex assembly sequences and so on. Most of these activities depend on general or ‘weak’ methods, primarily search. AI also addresses issues related to the representation and use of the knowledge of human experts. This unit of study will explore topics from selected areas of AI. Students who complete it will have an understanding of some of the fundamental methods and algorithms of AI, and an appreciation of how they can be applied to interesting problems. The unit of study will involve a practical component in which some simple problems are solved using standard AI techniques.The topics covered include intelligent agents, problem-solving and search, game-playing, machine knowledge-based systems, machine learning and neural networks.

The alternative advanced unit COMP3902 covers all the material of COMP3002, plus extra topics .The two units share the same lectures, but have different tutorials and assessment.
We hope you will enjoy this course!

Timetable

Lectures: Tuesdays 10-11am and 12-1pm, Carslaw Lecture Theatre 275
Tutorials (start week 2), location: Storie Dixon 432A
 
Tutorial
Time
Tutor
A (adv.) Wed 9-10am Daren Ler
B (adv.) Wed 10-11pm Irena Koprinska
C Wed, 12-1pm Harry Mak
D Wed, 1-2pm Clinton Freeman
E Fri, 10-11am John Drake
F Fri, 11-12noon John Drake
G Fri, 12-1pm Clinton Freeman

Resources

A Message board is available. Its aim is to allow you to communicate with other students on issues regarding this course. John Drake, the teaching assistant for this course, will check it regurlarly (every 2 or 3 days) and provide answers if necessary, that is when they can't be answered by other students. If your request is extremely urgent and you can't obtain suitable information from other students, it is preferable to send us an email.

Recommended books

1. S.J. Russell and P.Norvig, Artificial Intelligence, A Modern Approach
Prentice Hall, 0-13-103805-2, 1995

2. Ian H. Witten, Eibe Frank
Data mining - practical machine learning tools and techniques with Java implementations,
Morgan Kaufmann. 1-55860-552-5, 2001

Both are available at the Co-op bookshop and are also put on special reserve at the library.

Software resources
WEKA
Matlab Neural Network toolbox

Assessment Overview

Your final mark for the course will be calculated according to the following:
  • Weekly exercises:  9 % (3 weeks will be randomly marked for 3 marks each)
  • Assignments: 26% (Assignment 1 - 12% and Assignment 2 - 14%)
  • Exam: 65% (with a minimum of 26 required to pass the course, i.e. 40% on the exam)


    • About plagiarism: please do not confuse legitimate co-operation and cheating.
    Plagiarism or any other form of dishonesty will be dealt with by standard University procedures. Please also see the School Policy on Academic Honesty.

    Assignments

    Assignments can be done in pairs.
    Submission: 1) hard copy (report+code) in a folder to your tutor at the beginning of your tutorial class and  2) electronically (only the code) by e-mail or netfile to your tutor.
    Assignment 1 ( [regular] , [advanced] , [input-file]): out week 3, due week 6
    Assignment 2 ( [regular] , [advanced] ): out week 8, due week 11.

    Late submission of assignments policy:
    - a penalty of 1 mark per each day after the deadline will apply
    - assignments will not be accepted if the delay is more than 7 days

    If you have a condition requiring a special consideration, you must: 1) submit a form to the School office within 1 week from the date when assessment was due, 2) include your e-mail address, phone number and the name of your tutor, and 3) e-mail your lecturer that you have submitted a special consideration form.
    See the School Policy on Special Considerations due to Illness or Misadventure.

    Exam

    You will be allowed one A4 sheet with hand-written or typed notes (you can use both sides) and a non-programable calculator. No other material is allowed (no book, no additional notes). It will be a 2-hour exam, plus 10 mns reading time.

    Important note: Students who score less than 26 at the exam will fail the course regardless of their homework and assignment marks.
    Sample exam paper will be available in week 13.

    Lecture notes

    week Topic
    1 Introduction [pdf-3sld , pdf-6sld] 
    Intelligent agents [pdf-3sld , pdf-6sld] [ans]
    2 Problem solving and search. [pdf-3sld , pdf-6sld] 
    Informed search 1. [pdf-3sld , pdf-6sld] 
    3 Informed search 2  [included in the previous lecture notes] 
    Genetic algorithms. (with Dr Josiah Poon)  [pdf-3sld , pdf-6sld]
    4 Game-playing.[pdf-3sld , pdf-6sld] 
    5 Introduction to machine learning. [pdf-3sld , pdf-6sld] 
    Inferring rudimentary rules. Instance-based learning. [pdf-3sld , pdf-6sld] 
    6 Decision trees. [pdf-3sld , pdf-6sld , pdf-1sld] [add] 
    3 & 6 slides per page handouts - there are some problems with the layout of the formulas  on some of the slides; 1 slide per page  is provided as well - no problems with the formulas 
    7 Naive Bayes. [pdf-3sld , pdf-6sld] 
    Experimental evaluation of learning algorithms.  [pdf-3sld , pdf-6sld]
    Ensembles of learning algorithms. 
    8 Introduction to neural networks. [pdf-3sld , pdf-6sld] 
    Perceptron. [pdf-3sld , pdf-6sld]
    9 Backpropagation. [pdf-3sld , pdf-6sld] 
    Note: on slide 23 the value of o_3 was incorrectly copied from the previous calculations; it should be 0.53 not 0.65 => delta_3=-0.018 not -0.0019. The lecture notes have been corrected.
    10 Clustering. Self-organizing feature maps.  [pdf-3sld , pdf-6sld] 
    Learning vector quantization. 
    11 Knowledge representation 1: knowledge, semantic networks, frames and rule-based systems 1.  [pdf-3sld , pdf-6sld] [ans] 
    12 Knowledge representation 2: rule-based systems 2, expert systems.  [pdf-3sld , pdf-6sld] [mycin]
    Revision and preparation for the exam.  [pdf-3sld , pdf-6sld] (updated, contain the answers)
    Sample exam paper 
    13 Natural language processing (with Prof. Jon Patrick) 

    Tutorial exercises

    Attendance at your assigned tutorials is very important. If you want to change to another time, you need to see the Help Desk and ask to have the change made officially.

    Each week you must hand in to your tutor the exercises that are marked in the table below. These weekly exercises are to be done in pairs. They are the straight application of the material seen during lectures. They are easy and and their aim is to help you prepare better for the tutorial class. The other exercises will be done during the tutorial class.

    Please note that homeworks should be clearly presented. Messy and unreadable homeworks will not be assessed.
     
    week
    		 Tutorial Exercises Exercises for Homework
    2 [pdf]  All students: ex. 1c and 2a, c, d 
    Note: ex 10 (adv), this is 3.13 from R.&N
    "We said that at least one direction of bi-directionsl search must be a BFS...". This is not correct.  One direction must store the nodes generated, so that the other one can look for them. So, any search algorithm that stores all nodes in memory can be used, not necessarily BFS.
    3  [pdf]  Regular students -  ex. 3 
    Advanced - ex. 3 and 6 
    4  [pdf]  All students - ex. 4
    5  [pdf]  All students - ex. 2
    6  [pdf]  No homework exercises - assignment 1 is due 
    7   [pdf]  All students - ex. 1, 2, 6
    8   [pdf]  All students: ex. 1, 2 
    Ex.2: When training, apply the examples in the given order
    9   [pdf]  Regular students: ex. 1 
    Advanced - ex. 1 and 2
    10  [pdf]  All students - ex. 1 and 3
    11  [pdf]  No homework exercises - assignment 2 is due
    12  [pdf]   All students - ex. 1
    13  [pdf]  No homework exercises

    Tutorial solutions will not be put on the web. A hard copy will be available from your tutor at the following tutorial session.

    Check your marks!

    Your marks can be checked here. It is important that you double-check your marks throughout the semester to avoid chasing them up with your tutors after the end of the semester. If you notice any mark missing or incorrect report it as soon as possible to your tutor (not to the lecturer).
     

    Contact details

    Dr Irena Koprinska
    Email: irena@it.usyd.edu.au
    Office: Madsen Building, G90A
    Consultation time: Tuesdays 1-2pm

    Last updated: 6 June 2003