146-148 article

so from the article as one of the representative learning methods, has shown to achieve superior performance in many applications. Therefore, the more attention and variants of Auto Encoders have been reported including Contractive Auto-Encoders, Denoising Auto-Encoders, Lightweight Encoding and Nonnegative Constraints Instability Auto Encoders. Recently, the Auto Discriminatory Encoder is reported to improve performance by taking into account in the classroom and among the classroom information. In this paper, we propose Large Margin Auto-Encoders to further boost discrimination by applying different sample classes to be distributed on a large scale in the hidden feature space. In particular, we accumulate a single layer LMAE to build a deep neural network to learn the right features. And finally we put this feature into the classifier classmate softmax. Extensive experiments were performed on MNIST datasets and CIFAR-10 datasets for each classification. The experimental results show that the proposed LMAE exceeds the traditional Auto-Encoders algorithm.