Mnist4

Mnist has a training set of 10000 examples, and the first four categories(0-3) of it, i.e., the first 4000 samples form a subset called Mnist4. It is divided into 4 categories and described by 3 different features: 30 Isometric Projection (IP), 9 Linear Discriminant Analysis (LDA) and 30 Neigh-bourhood Preserving Embedding (NPE).