A multiclass classification method by distance mapping learning network

We propose a method of multiclass classification by utilizing a distance mapping learning network that is a distance-based multilayer perceptron The network can obtain the non-linear mapping between the input objects and the outputs by providing a pair of objects and the desired distance between them. It thus realizes multiclass classification based on pairwise classifications iteratively. We show the validity of the model with two classification problems: Iris classification and facial expression classification.