How Did AQ Face The East-West Challenge? An Analysis of the AQ Family's Performance in the 2nd International Competition of Machine Learning Programs
Eric Bloedorn, Ibrahim F. Imam, Kenneth A. Kaufman, Marcus A. Maloof, Ryszard S. Michalski, Janusz Wnek · 1995
The “East-West Challenge” is the title of the second international competition of machine learning programs, organized in the Fall 1994 by Donald Michie, Stephen Muggleton, David Page and Ashwin Srinivasan from Oxford University. The goal of the competition was to solve the “TRAINS problems”, that is to discover the “simplest” classification rules for train-like structured objects. The rule complexity was judged by a Prolog program that counted the number of various components in the rule expressed in the from of Prolog Horn clauses. There were 65 entries from several countries submitted to the competition. The GMU team’s entry was generated by three members of the AQ family of learning programs: AQ-DT, INDUCE and AQ17-HCI. The paper analyses the results obtained by these programs and compares them to those obtained by other learning programs. It also presents ideas for further research that were inspired by the competition. One of these ideas is a challenge to the machine learning community to develop a measure of knowledge complexity that would adequately capture the “cognitive complexity” of knowledge. A preliminary measure of such cognitive complexity, called C-complexity, different from the Prologcomplexity (P-complexity) used in the competition, is briefly discussed.