Machine Learning - Summer 2003
Project 1 - Representing examples and hypotheses
Posted: 05/27/2003
Due: 06/02/2003
Do the following. For each item include a proper documentation in
your report:
-
Create a data file in Prolog format using the set of examples for the target
concept "PlayTennis", shown in Mitchell - Table 3.2 (page 59). Use the
following template (for an example of this format see the data file animals.pl):
example(ID,
Class, List_of_attribute_value_pairs). ID must be a
number.
Test the file for eventual syntax errors by loading it in Zprolog. Include
a printout in your report.
-
Using the covering.pl program and
specifically the predicates generalize(E,H) and model(H,M),
generate all hypotheses that cover more than one example from class
"no" and no examples from class "yes". Do all this
with a single query and include a copy of the query and the Zprolog output
in your report. Hint: see item #9 (second query) in Lab
Experiments 1.
-
Using the covering.pl program generate
pairs of different examples from class "no" and then apply lgg(E1,E2,H)
and model(H,M), to generate all hypotheses that cover examples
from class "no" , but do not cover examples from class "yes".
Hint: see query #10 in Lab
Experiments 1.
-
Using the above results create (manually) the following two sets of rules
(hypotheses):
-
a subset of the rules generated with generalize (item 2 above) that
jointly cover all examples from class "no".
-
a subset of the rules generated with lgg (item 3 above) that jointly
cover all examples from class "no".
Documentation and submission: Include the results of all the
work you did according to the above specifications in a document (Word
or HTML format) with a title page including your name and either send it
as an e-mail attachment or hand it as a hard copy.