COSC 575: Machine Learning

Project 1
Spring 2012

Due: Wed, Feb 1 @ 8 PM
10 points

1. Write a program to read, parse, store, and output examples in Mark's File Format (mff). Consider the following version of the Bikes data set:

   Make	Tires	Handle Bars	Water Bottle	Weight	Bike Type
   Trek	Knobby	Straight	y	250.3	Mountain
   Bridgestone	Treads	Straight	y	200	Hybrid
   Cannondale	Knobby	Curved	n	222.9	Mountain
   Nishiki	Treads	Curved	y	190.3	Hybrid
   Trek	Treads	Straight	y	196.8	Hybrid

   The file bikes.mff shows this data set in Mark's File Format. Files containing valid data sets begin with '@dataset' followed by an identifier. Attribute declarations appear next. The string '@attribute' precedes each declaration, which is for a symbolic attribute or a numeric attribute. The attribute's name appears next, followed by its domain. The domain for symbolic attributes is a list of values separated by whitespace. The domain for numeric attributes is not explicitly specified and is assumed to be the set of representable floating-point numbers.

   The token '@examples' separates the attribute declarations from the examples, which are simply values separated by whitespace.

   For simplicity, you can assume that all elements of the file are separated by at least one space character. Moreover, attribute declarations and examples will appear on single lines. I have defined a grammar for the file format.

2. In subsequent projects, operations on the data set you will need to perform include
   • identify illegal attribute values
   • iterate over examples
   • iterate over attribute values
   • execute logic based on the attribute's type

   I would recommend implementing two primary structures: one for the information about attributes, and one for the examples themselves. I have produced design document for the classes and their methods. You do not have to follow my design exactly, but I'm reasonably certain that my design contains all of the essential functionality for the subsequent projects.

3. The program should take input from the command line. Use -t to specify the name of the training file and use -T to specify the name of the testing file, if any.

   % java p1 -t bikes.mff
   % java p1 -t bikes.train
   % java p1 -t bikes.train -T bikes.test
   % java p1 -t bikes-tr.mff -T bikes-te.mff
   

   The program should perform light checks for proper formatting and data integrity. For this project, the program can simply read the input files, populate the appropriate objects, and output the examples to the console.

4. Implement the program using the language of your choosing (e.g., ANSI C, ANSI C++, Java, Fortran 77, Lisp, ruby 1.8, python 2.6); however, it must compile and run on my Unix machine. The implementation must be general, meaning that it should work for all possible data sets. You must implement your program using the standard libraries the language provides. If you want to use external resources or something non-standard, check with me first. Finally, your project must be easy to build and run, and you should include instructions in a README file that describes exactly how to build and run your project. This is especially true for people who use Java.

Instructions for Submission

//
// Name
// E-mail Address
// Platform: Windows, OS X, Linux, Solaris (daruma), etc.
// Language/Environment: gcc, g++, java, g77, ruby, python, haskell, etc.
//
// In accordance with the class policies and Georgetown's Honor Code,
// I certify that, with the exceptions of the class resources and those
// items noted below, I have neither given nor received any assistance
// on this project.
//