ComS 573:
Machine Learning
Department of Computer Science
Iowa State University
Spring 2008
Laboratory Assignment 3
Due: Wednesday March 26, 12pm noon.
In this assignment, you will experiment with the Neural Network classifier.
Data
-
Optical Recognition of Handwritten Digits (use optdigits.names,
optdigits.tra as training data, and optdigits.tes as test data) from the
UC Irvine Machine Learning
Repository.
Tasks
- Test the neural network classifier on the dataset. Report the performance
of the trained networks by computing the classification accuracy over the
test data (optdigits.tes). You may want to use the
validation method as the stopping criterion (say using 1/3 of training data
optdigits.tra as the validation set).
- Experiment with different number of hidden
units, different learning rates and momentum rates.
Compare their convergence speed and performance.
What to Turn In
Turn in:
- A report (in electronic form) of the experimental
results and discussions of these results. It should include a brief
description of all the design choices made. Try to use tables or plots to
summarize your results.
You should specify the parameters
of every experiment in such a way they can be replicated by the T.A., and justify your decision about the use (or not) of these
parameters.
- Any source code that you may have written.
- The data file in arff format.
Submission instructions
You should submit your answers via email.
Compress the source code of your program (if you wrote some), the arff files, and the report of the results,
and email the compressed (zip, rar, tar) file to rjordan at iastate.edu (change at by @) before
the deadline.
You will receive a confirmation email before 12:10 p.m. (noon).