CS2411 Subsymbolic Artificial Intelligence and Neural Networks

2004-2005

FAQ| Announcements| Course Timetable| Handouts| Lab Information| Example Class Sheets| Reading Lists| Links

Ask a Question


Announcements

11/1/05 - A sheet giving some exam advice is available.
30/9/04 - Monday lab and examples class sessions will not run, (unless class size increases greatly).
30/09/04 - Welcome
Course starts today.


Lab and Example Class Timetable

Course Handouts

 
First day handout:
General information about the course
Neural Networks I:
General aspects of neural networks and simple perceptrons, as
gzipped Postscript or as PDF.
Generalization and Testing :
Elements of machine learning, and how to test a learned system. As
gzipped Postscript or as PDF.
Neural Networks II:
Multi-layer perceptrons and gradient descent learning. As
gzipped Postscript or as PDF.
Slides from a lecture on applications of MLPs
as PDF .
Neural Networks III:
Improving performance of neural networks. As
gzipped Postscript or as PDF.
Probability I:
A review of probability and Bayes Rule.
as gzipped Postscript or as PDF .
Bayesian Classification:
Bayesian Classification and Probability Estimation.
as Postscript or as PDF .
Document Classification:
Bayesian classification as applied to document classification. Also includes a section on on-line learning of probability estimates.
as Postscript or as PDF .
Decision Theory:
Given the probabilities of the classes, what is the best classification to make. Includes discussion of risk, and ROC curves.
as Postscript or as PDF .
Search I
The general problem of search, greedy search and hill-climbing as postscript or . as PDF .
Search II
Genetic Algorithms as postscript or . as PDF .


Lab Information

Matlab Tutorial:
Matlab Tutorial in PDF or Matlab Tutorial in Postscript.
This tells gives an introduction to Matlab. Work through this before the first lab. There are problems in the tutorial, but they are not a part of the lab; they are just to help you learn Matlab. They are optional.
Lab 1: Introduction to Matlab and Classification
Description of Lab 1
Lab 2: Classification using Multilayer Perceptrons
Instructions for Lab 2
Lab 3: Bayesian Spam Filter
Commands for Lab 3 in PDF
Note: Lab 3 is a two session lab. Hints for Lab 3
Lab 4: Genetic Algorithms
Commands for Lab 4 in PDF


Examples Classes

Questions

  1. First Examples Class questions This is a sheet of the questions only.
  2. Second Examples Class questions
  3. Third Examples Class questions
  4. Fourth Examples Class questions
  5. Fifth Examples Class questions

Answers

  1. Answers to the first Examples Class questions as PDF or as postscript .
  2. Answers to second Examples Class questions
  3. Answers to third Examples Class questions
  4. Answers to fourth Examples Class questions
  5. Answers to fifth Examples Class questions


Reading Lists by Topic

A reading list by topic is here.


Links

A page on Multi-layer Perceptrons
from Jim Marshall at Pomona College. Includes backpropagation and applications.
An Introduction to Neural Networks. An on-line book by Krose and van der Smagt.
Look at section 3.1 and 3.2 on perceptrons, and chapter 4 on backpropagation (but we need not worry about the mathematical details of back propagation. Focus on the performance details.)
Ant Colony Optimization Applet
I mentioned ant colony optimization for the TSP in the first lecture as an example of the subsymbolic approach to AI. Here is a link to a Applet created by Mark C. Sinclair showing this running on a very simple problem.
Elastic Net Applet
The elastic net was mentioned in the first lecture as an example of the subsymbolic approach to AI. Here is a link to a Applet created by Alexander Budnik and Tanya Filipova showing this running on a very simple problem.