CS575, Artificial Intelligence II
Spring 1999

Assignment 4: Neural Networks

Object

To understand the techniques behind the standard techniques for learning using a feedforward, backpropagation neural network, and to use such a network on real data.

Domain

We will use recognition of handwritten characters as the domain. The task will be to distinguish handwritten versions of a single character from other characters and/or images.

Procedure

Write a program the uses the backpropagation code in ~rth/public/bpnn. Use the same attribute data that your used for the decision tree assignment, and train a network to distinguish between your chosen handwritten character, and any others. Look at these notes to help you with the code. Split  your examples into two sets as befoer for training and testing.

Deliverables

  1. A printout showing how the network learns. A summary of error values for each epoch is a good way to do this.
  2. A printout showing how good the network is on the test cases. Again a summary of errors is necessary.
  3. A print of the program code.

Due date

Wednesday May 12th, before 5pm.