Multi-class boosting based on Phase-out Model

We present a probability model for multi-classification problem. In our theory, a formalization description method in the form of probability is put forward for multi-class boosting. With these arguments, a novel prediction framework, called Phase-out Model, is proposed. Unlike previous classifiers which choose the best one from all classes as the prediction, our model weeds out one of all classes step by step until only one still insists when predict and the surviving class is the outcome. A local optimum algorithm is designed to implement the model. Experiments show that our algorithm is feasible and our model is more robust than traditional prediction framework.