CS4341 Introduction to Artificial Intelligence
Project 5 - D 2001
Due Date:
Monday, April 23, 2001 at 5:00 pm.
PROJECT DESCRIPTION
Construct a learning system for face recognition using neural networks
and the error back propagation procedure. This project is based on the
source code and dataset provided online as a companion to Chapter 4 of
Tom M. Mitchell's "Machine Learning" textbook (McGraw-Hill, 1997).
For your convenience, here is a PDF version of
the code documentation.
Also, thanks to Ben Holt for putting together, and for making available
to the rest of the class,
a text version of the assignment and a
tarball
fixing the training and test lists to use relative paths, and a few
other minor improvements.
PROJECT ASSIGNMENT
This project consists of two parts:
- Face Recognizer.
Train a neural network to recognize who the person in a picture is
among a group of 20 possible people.
- Pose Recognizer.
Train a neural network to recognize whether the person in a picture is
looking up, straight, left, or right.
You must follow the guidelines below for the training of your neural nets:
- Assignment: Complete Part I of the assignment described in the
data and code webpage (under "Documentation") by Tom Mitchell.
Include answers to the all the questions in that assignment in your written report.
For your convenience, here is a PDF version of
the code documentation and of Mitchell's assignment.
- Code: You can use
Your code must run on the CS or CCC Unix machines.
Make sure that the code you use correctly computes the error function of
the neural net when more than one output node is used.
For that you need to read the code and understand it in detail.
- Training/Test Instances:
Use the "quarter"-size collection of pictures provided at the
CMU site. You can find copies of those pictures in our course
public directory /cs/cs4341/Proj5/ on the
CCC machines.
REPORT AND DUE DATE
Project 5 is due on Monday, April 23 at 5:00 pm.
Your system should follow the
Departmental Documentation Standard.
The submission should be done using the
turnin program.
- Program and Neural Networks.
You should submit (1) the source code of your program (only if you developed
your own) and (2) the neural networks that you obtained.
- Written Report.
Your report should discuss the following issues:
- answers to all questions in Part I of Mitchell's assignment
(for your convenience, here is a PDF version of
the code documentation and of Mitchell's assignment),
- adaptation of the code (if any) or a description of your own code,
- the experiments you ran with the system,
- the topology (number of units in each hidden layer),
initial weights, number of iterations of the error backpropagation
algorithm, and final weights of each of your neural nets,
- evaluation of the accuracy of each of the neural nets,
- strengths and the weaknesses of the system.
Your report should also include a short user manual explaining how to
install, run, and use your system (if different from the CMU package).
GRADUATE CREDIT
No additional graduate problem is required for this project.
Just do your best on the project described above!