Multi-class classifier-based adaboost algorithm

Abstract

A multi-class classifier-based AdaBoost algorithm for the efficient classification of multi-class data is proposed in this paper. The traditional AdaBoost algorithm is basically a binary classifier and it has limitations when applied to multi-class data problems even though its multi-class versions are available. In order to overcome the problems of the AdaBoost algorithm for multi-class classification problems, we devise a AdaBoost architecture with its training algorithm that uses multi-class classifiers for its weak classifiers instead of series of binary classifiers. The proposed AdaBoost architecture can save its training time drastically and obtain more stable and more accurate classification results than a typical multi-class AdaBoost architecture based on binary weak classifiers. Experiments on an image classification problem with collected satellite image database are preformed. The results show that the proposed AdaBoost architecture can reduce its training time 50%- 70% depending on the number of training rounds while maintaining its classification accuracy competitive when compared to Adaboost.M2. © 2012 Springer-Verlag.