Inducing compact NNTrees using discriminant rough null space method

A Neural Network Tree (NNTree) is a hybrid learning model. NNTrees are more suitable for structural learning and can make decisions faster than normal neural networks. The goal of this research is to embed the NNTrees into different portable devices. To reach this goal, it is necessary to induce compact NNTrees that can be implemented easily on a chip. So far, we have tried several dimensionality reduction approaches, including principle component analysis (PCA), linear discriminant analysis (LDA), direct centroid (DC) approach, and discriminative multiple centroid (DMC) approach. In this paper, we investigate the discriminant rough null space (DRNS) approach.