CSCI 4150 Introduction to Artificial Intelligence 2009

◇◇◇

... Instructor: Dr. Thomas Trappenberg ... Office: room 313 .... email: tt@cs.dal.ca .... Office hour: most times I'm in my office or send email ...

Grading schemes:

60% Assignments, 40% Project

Recommended textbook:

Russell & Norvig, Artificial Intelligence: A modern approach, Prentice Hall

Lecture notes by Andrew Ng: http://www.stanford.edu/class/cs229/materials.html

Resources:

An often used implementation of SVMs with MATLAB interface is LIBSVM

Plegarism:

Please familiarize yourself with the university policy on Intellectual Honesty. Every suspected case will be reported.

Lectures:

January 6: Introduction, different areas of AI, history, AI1.pdf

January 8: Agents, Brief intor to Matlab

January 13: Uninformed search, AI2.pdf

January 15: Heuristic search, AI3.pdf (update Jan 16), creature_state function

January 20: Local beam search, GA, hill climbing, simulated anealing, AI4.pdf

January 22: gradient descent, regression problems, Ng1.pdf

February 5: Generative learning algorithms, Ng2.pdf

February 10: SVM, Ng3.pdf, Burges98.pdf, Berwick03.pdf

Assignments:

1. Missionaries and Cannibals problem
2. Give the data set for a two dimensional linear model, estimate the parameters of the model with gradient descent and the normal methods using Matlab. Comment on the different algorithms.

   Due Tuesday, January 27, 2009. Send email to tt@cs.dal.ca with subject line AIa2, or bring printout to class.

3. a) Demonstrate and study the performance of the classification algorithms logistic regression, Gaussian discriminant analysis, and Naive Bayes on the following examples. Data for two class should be chosen from a Gaussian distribution with a diagonal covariance matrix of unit strength. The center of one class is (1,1), the other is (2,2), and both classes are equally likely. Generate a training set for training the classifiers. The performance of the classifier should be tested with an independent test set. Discuss you results briefly.
   b) Repeat the analysis for the case when one of the classes is twice as likely as the other.

   Due Thursday, February 12, 2009. Send email to tt@cs.dal.ca with subject line AIa3, or bring printout to class.

4. Study the spiral classification problem given by the image spiral.bmp with a SVM and a generative model. Discuss which of the methods is working better on this problem.

   Note: You can use LIBSVM (see resource link above) or any other SVM implementation. To read an image file into a Matlab array, you can use the command imread('spiral.bmp').

   Due Tuesday, March 3, 2009. Send email to tt@cs.dal.ca with subject line AIa4.

Course Project

This project should be made with a partner. So, please team up and let me know in case you did not find a partner.

The purpose of the project is to demonstrate and discuss the application of machine learning algorithms to a specific application. Chose a data set from the UCI machine learning repository, preferably one with a classification problem. Study your chosen problem with some of the algorithms which we discussed in class. You might want to consult the literature that might exist for this problem, and you should consult with the course instructor about your progress.

The results of your studies should be written up in form of a scientific paper. Some introduction to scientific writing will be given later in class, and the paper has to comply with strict guidelines given below.

Paper guidelines

The paper must be submitted by email to tt@cs.dal.ca before April 8, 2009, with the subject line `AIproject'. Please use letter format with font size of at least 11pt. The page limit is 6 pages including possible appendices, references and figures. The paper must include a title, the name and affiliation of the authors, and an abstract not exceeding 100 words. References must comply to APA style.

Brief guideline and tips for writing a scientific paper