ICS 361 -- Artificial Intelligence Programming

Machine Learning

Assignment #4

This assignment is worth 150 points.

Read the article written by Tom Mitchell titled The Discipline of Machine Learning, posted on Laulima.

  1. (75 p) Luger's supervised ID3 machine learning code.
    (20 p) Load Luger's code with the credit example data (ID3-luger and ID3-credit-data). Note that you may encounter a name conflict when loading the files into Lisp. If so, modify the name(s) of the conflicting functions so that the code runs as specified. The code can be found on Laulima.
  2. (75 p) Mooney's COBWEB code for clustering (unsupervised learning).
    (20 p) Load the main files: ml-utilities.lisp and cobweb.lisp. Load each of the four data files: cobweb-animal-dnata.lisp, cobweb-animal2-data.lisp, cobweb-soybean-data.lisp, cobweb-soybean2-data.lisp. The code can be found on Laulima.
  3. (15 p) Extra credit - Create (or find (cite sources)) a data set in a NEW domain/task for either ID3 or COBWEB. Create at least 15 instances in your domain with at least 7 properties in each instance. Run experiments and describe in detail the results of the classifier/clustering algorithm on the data in your new domain.
  4. (25 p) Extra Credit - Find another learning program (not in the style of ID3 or COBWEB, e.g. neural networks). Load and run the code on the examples provided. Test the program on a new/novel set of data. Report on how this method of machine learning is different and better or worse that ID3 and COBWEB. Include the type of problems for which it is best suited when compared with other styles of learning. What are the constraints on the data?

Turn in the following:

Plain text files containing your code and scripts. A report with your results and discussion in a standard format like rtf, pdf, or text.

(c) N. E. Reed, 2004-2016