Robust Bidding in Learning Classifier Systems Using Loan and Bid History
Abrham Workineh
Abdollah Homaifar
Autonomous Control and Information Technology Center
Department of Electrical and Computer Engineering
North Carolina A & T State University
Greensboro, NC 27411
atworkin@ncat.edu, homaifar@ncat.edu
Abstract
In this paper, we introduce bid history and loan concepts to mitigate the shortcomings of the bidding strategy in traditional learning classifier systems (LCSs). In direct analogy with real auctions, all classifiers matching the current input compare the average bid history with their potential bid based on their current strength. The average bid history parameter gives general information about the auction (potential of competent classifiers) and determines the minimum loan amount a classifier should request. Debt and due date parameters have also been added to the traditional LCS parameter list to keep track of the transaction status, accuracy, and experience for granting or denying loan requests. The results obtained show a significant improvement on the convergence of the learning system.
