Multiclass Cascades for Ensemble-based Boosting Algorithms

We propose a general method applicable to existing multiclass boosting-algorithms for creating cascaded classifiers. The motivation is to introduce more tractability to machine learning tasks which require large datasets and involve complex decision boundaries, by way of separate-and-conquer strategies that reduce both the training and detection-phase overheads. The preliminary study explored the application of our method to AdaBoost. ECC on six UCI datasets and found that a decrease in the computational training and evaluation overheads occurred without significant effects on the generalization of the classifiers.