Optimal dimensionality of metric space for classification

摘要: In many real-world applications, Euclidean distance in the original space is not good due to curse of dimensionality. this paper, we propose a new method, called Discriminant Neighborhood Embedding (DNE), learn an appropriate metric for classification given finite training samples. We define discriminant adjacent matrix favor task, i.e., neighboring samples same class are squeezed but those different classes separated as far possible. The optimal dimensionality can be estimated by spectral analysis proposed which great significance high-dimensional patterns. Experiments with various datasets demonstrate effectiveness our method.