Integration of Learning Based Boosting and Programming Based Boosting for Multiple Classifier Classification in Biometric Based Person Authentication

The present paper proposes the integration of Programming and Learning Based Boosting which is more reliable than either Programming or Learning Based Boosting because it combines the advantages of Learning based Boosting and Programming based Boosting while removing the drawbacks of both of them. Thus, unlike assigning the weights of individual classifiers and keeping those weights hard fixed all throughout (in case of Programming Based Boosting) or arbitrarily and randomly assigning the initial weights of individual classifier (in case of Learning Based Boosting) this method initially fixes up the weights of the individual classifier according to their performance evaluation (accuracy) and then updates the different weights according to one weight learning algorithm methodology. In this way, this method removes the drawbacks of both Programming and Learning Based Boosting Method while retains the advantages of both or them.