Constructing discriminative biorthogonal bases for classification

We present a method for constructing discriminative biorthogonal bases for classification. In the proposed method we use idea of the lifting scheme a method for constructing biorthogonal wavelets and Support Vector Machines used for building classifiers. Combining those two ideas resulted in the construction of classifiers which are based on information selected form the available data and still yield very good classification accuracies. We have shown that our method can be treated as a simple feature extractor for other classification algorithms or as a method that produces a set of classifiers whose outputs may be combined to give final classification.