Multistage building learning based on misclassification measure

In the areas of machine learning and pattern recognition, discriminative learning methods are well-known for giving better classification performance than the methods which estimate probabilistic distributions of data. In this paper, we propose a framework of multi-stage classification based on the minimum classification error/generalized probabilistic descent learning which is one of the promising discriminative learning methods. The proposed method makes it possible to use misclassified data to improve the classification performance by incorporating the supplemental features in the original feature vector space