HarryPoter

**A Study of Deep Learning Methods for Image Classification** We address the question of how deep neural networks (DNNs) model the data in ways that are not typical for DNNs, and explore the possibilities. The deep networks are typically based on learning to extract the image information from the training data, that is, in a binary or mixed representation, but the training data are often labeled by a vector representation and then the labels can be learned by applying deep …