Prerequisites: Graduate standing, knowledge of programming, three semesters of calculus.

Instructor: Dr. Anthony Maida.

Requirements and grading:
Midterm (take home):	30%
Final (take home):	35%
Projects and assignments:	35%

Due Dates 2007:
Sept 13:	First proposal draft due (1 page)
Sept 27:	Second proposal draft due (2 pages)
Oct 23:	Midterm take home exam given out
Nov 1:	Midterm exam due (30%)
Nov 29:	Final project due (35%)
Nov 29:	Take home cumulative final given out
Dec 6:	Take home final due (35%)

Textbook:

Neural Networks for Pattern Recognition

Christopher M. Bishop

Clarendon Press, Oxford, 1995

ISBN: 0-19-853864-2

It helps to prototype simple neural networks in MatLab. A MatLab tutorial is available here.

A small MatLab back propagation program for XOR is here.

• Instructions to build a backprop net in Java for computing XOR are here.
• Instructions to build a novelty detection net are here.
• Instructions to build a synchronization grid of integrate-and-fire neurons are here.
• Link to a good site on Neural-network-based object recognition is here.
• Reference code to play with Dr. Izhikevich's neural compartments is here in zip format. His web page is here.

• Artificial neural network overview slides are here.

• The chapter on early stages of vision is here.
• The chapter on pattern recognition is here.
• The chapter on associative memory is here.
• The chapter on recruitment learning is here.

Bishop Lecture Notes: Compressed files of my scanned lecture notes from the Bishop text can be downloaded here. When uncompressed they are jpeg files. Topics follow:

Chapter 1

Chapter 2

Chapter 3

The data for the Monks One problem is given below.

For the training data, the input features are here and the desired output is here.

For the testing data, the input features are here and the desired output is here.

Here is a link to a Java-based tutorial on neural networks.

Sutton and Barto's online book on reinforcement learning.

A link othat uses reinforcement learning and neural networks for motor control.

This is a link to Gabor filter implementations in Matlab to be used for visual feature preprocessing.

The source code for sample project is on this file: Psearch.tgz. Your browser my uncompress the file automatically. Depending on the extension of the downloaded file use either the command: "tar -zxf Psearch.tgz" or the command : "tar -xf Psearch.tar". It should create a directory named PsearchExportTest with the source files.

If you have trouble uncompressing the "tgz" file above, here is a zip file: Psearch.zip. You should be able to expand it with the "unzip" command.

Documentation for the above is here and here.