A Study on Deep Learning Architecture and Their Applications

Abstract

Deep learning (DL) techniques are currently drawn more research interest including computer vision, pattern recognition, speech recognition, and robotics due to its higher accuracy and self-feature extraction property compared to the traditional algorithms that necessitate hand-crafted features. The success with the DL is due to the advancement in technology and the availability of the high-power devices that are achieved with high computational complexity. In this paper, we aim to provide an intrinsic investigation of the DL architectures and their applications in the practical world. Specifically, the overview of autoencoder, restricted Boltzmann machine, generative adversarial network, and convolutional neural network are provided. Different aspects and applications in real-world cases are surveyed and summarized.