Abstract
Deep learning is one strategy for generating feature representation. It maps input to output with multiple layers. This structure allows the model to learn the representation of complicated input as hierarchical concepts, building each concept upon each other from low-level features to more abstract representations. Thanks to the increase of the number of training examples and the improvement of computer hardware and software, deep learning has become more useful and powerful. In recent years, deep learning has demonstrated significant success in the field of computer vision, speech recognition, and natural language processing.