| | |
Summary: Al-Alaoui Pattern Recognition Algorithm: A MSE
Asymptotic Bayesian Approach to Boosting
Mohamad Adnan Al-Alaoui
Professor, ECE Department
American University of Beirut, Beirut, Lebanon
Abstract - The relation of the Al-Alaoui pattern
recognition algorithm to the boosting and bagging
approaches to pattern recognition is delineated. It is
shown that the Al-Alaoui algorithm shares with bagging
and boosting the concepts of replicating and weighting
instances of the training set. Additionally it is shown
that the Al-Alaoui algorithm provides a Mean Square
Error, MSE, asymptotic Bayesian approximation to
boosting. Experimental results demonstrate the viability
of the Al-Alaoui algorithm for pattern classification.
.
Keywords: Back-propagation algorithm; bagging;
Bayes classifier; boosting; mean-square-error (MSE);
Neural networks (NNs).
1 Introduction
|
